SQUID Frequently Asked Questions © 2001 Duane Wessels, Frequently Asked Questions (with answers!) about the Squid Internet Object Cache software.

You can download the FAQ as , , , or as a .

About Squid, this FAQ, and other Squid information resources What is Squid?

Squid is a high-performance proxy caching server for web clients, supporting FTP, gopher, and HTTP data objects. Unlike traditional caching software, Squid handles all requests in a single, non-blocking, I/O-driven process. Squid keeps meta data and especially hot objects cached in RAM, caches DNS lookups, supports non-blocking DNS lookups, and implements negative caching of failed requests. Squid supports SSL, extensive access controls, and full request logging. By using the lightweight Internet Cache Protocol, Squid caches can be arranged in a hierarchy or mesh for additional bandwidth savings.

Squid consists of a main server program Squid is derived from the ARPA-funded . What is Internet object caching?

Internet object caching is a way to store requested Internet objects (i.e., data available via the HTTP, FTP, and gopher protocols) on a system closer to the requesting site than to the source. Web browsers can then use the local Squid cache as a proxy HTTP server, reducing access time as well as bandwidth consumption. Why is it called Squid?

Harris' Lament says, ``All the good ones are taken."

We needed to distinguish this new version from the Harvest cache software. Squid was the code name for initial development, and it stuck. What is the latest version of Squid?

Squid is updated often; please see for the most recent versions. Who is responsible for Squid?

Squid is the result of efforts by numerous individuals from the Internet community. of the National Laboratory for Applied Network Research (funded by the National Science Foundation) leads code development. Please see for a list of our excellent contributors. Where can I get Squid?

You can download Squid via FTP from or one of the many worldwide .

Many sushi bars also have Squid. What Operating Systems does Squid support?

The software is designed to operate on any modern Unix system, and is known to work on at least the following platforms: Linux FreeBSD NetBSD OpenBSD BSDI Mac OS/X OSF/Digital Unix/Tru64 IRIX SunOS/Solaris NeXTStep SCO Unix AIX HP-UX

For more specific information, please see . If you encounter any platform-specific problems, please let us know by registering a entry in our . Does Squid run on Windows NT?

Recent versions of Squid will / packages.

maintains the native NT port of Squid and is actively working on having the needed changes integrated into the standard Squid distribution. Partially based on earlier NT port by .

has ported Squid to Windows NT and sells a supported version. You can also download the source from . Thanks to LogiSense for making the code available as required by the GPL terms. What Squid mailing lists are available?

squid-users@squid-cache.org: general discussions about the Squid cache software. Subscribe via , and also at . squid-users-digest: digested (daily) version of above. Subscribe via squid-announce@squid-cache.org: A receive-only list for announcements of new versions. Subscribe via

We also have a few other mailing lists which are not strictly Squid-related. I can't figure out how to unsubscribe from your mailing list.

All of our mailing lists have ``-subscribe'' and ``-unsubscribe'' addresses that you must use for subscribe and unsubscribe requests. To unsubscribe from the squid-users list, you send a message to What other Squid-related documentation is available?

for information on the Squid software to be published by January 2004. gives information on our operational mesh of caches. (uh, you're reading it). . . Squid documentation in , , , , and another in . . Yeah, its extremely incomplete. I assure you this is the most recent version. ICPv2 -- Protocol ICPv2 -- Application Does Squid support SSL/HTTPS/TLS?

As of version 2.5, Squid can terminate SSL connections. This is perhaps only useful in a surrogate (http accelerator) configuration. You must run configure with Squid also supports these encrypted protocols by ``tunelling'' traffic between clients and servers. In this case, Squid can relay the encrypted bits between a client and a server.

Normally, when your browser comes across an The browser opens an SSL connection directly to the origin server. The browser tunnels the request through Squid with the

The and (expired). What's the legal status of Squid?

Squid is by the University of California San Diego. Squid uses some .

Squid is .

Squid is licensed under the terms of the . Is Squid year-2000 compliant?

We think so. Squid uses the Unix time format for all internal time representations. Potential problem areas are in printing and parsing other time representations. We have made the following fixes in to address the year 2000: timestamps use 4-digit years instead of just 2 digits.

Year-2000 fixes were applied to the following Squid versions: : Year parsing bug fixed for dates in the "Wed Jun 9 01:29:59 1993 GMT" format (Richard Kettlewell). squid-1.1.22: Fixed likely year-2000 bug in ftpget's timestamp parsing (Henrik Nordstrom). squid-1.1.20: Misc fixes (Arjan de Vet).

Patches: . If you are still running 1.1.X, then you should apply this patch to your source and recompile. . .

Squid-2.2 and earlier versions have a . This is not strictly a Year-2000 bug; it would happen on the first day of any year. Can I pay someone for Squid support?

Yep. Please see the . Squid FAQ contributors

The following people have made contributions to this document:

Please send corrections, updates, and comments to: . About This Document

This document is copyrighted (2000) by Duane Wessels.

This document was written in SGML and converted with the .

Most current version of this document can always be found at in HTML, Plain Text, Postscript and SGML formats. Want to contribute? Please write in SGML...

It is easier for us if you send us text which is close to "correct" SGML. The SQUID FAQ currently uses the LINUXDOC DTD. Its probably easiest to follow examples in the this file. Here are the basics:

Use the <url> tag for links, instead of HTML <A HREF ...> <url url="http://www.squid-cache.org" name="Squid Home Page">

Use <em> for emphasis, config options, and pathnames: <em>usr/local/squid/etc/squid.conf</em> <em/cache_peer/

Here is how you do lists: <itemize> <item>foo <item>bar </itemize>

Use <verb>, just like HTML's <PRE> to show unformatted text. Getting and Compiling Squid

You must download a source archive file of the form squid-x.y.z-src.tar.gz (eg, squid-1.1.6-src.tar.gz) from , or. . Context diffs are available for upgrading to new versions. These can be applied with the ). How do I compile Squid?

For % tar xzf squid-1.1.21-src.tar.gz % cd squid-1.1.21 % make

For % tar xzf squid-2.0.RELEASE-src.tar.gz % cd squid-2.0.RELEASE % ./configure % make What kind of compiler do I need?

To compile Squid, you will need an ANSI C compiler. Almost all modern Unix systems come with pre-installed compilers which work just fine. The old If you are uncertain about your system's C compiler, The GNU C compiler is available at . In addition to gcc, you may also want or need to install the What else do I need to compile Squid?

You will need installed on your system. Do you have pre-compiled binaries available?

The developers do not have the resources to make pre-compiled binaries available. Instead, we invest effort into making the source code very portable. Some people have made binary packages available. Please see our .

The site has pre-compiled packages for SGI IRIX.

Squid binaries for .

Squid binaries for

Gurkan Sengun has some available. How do I apply a patch or a diff?

You need the cd squid-2.5.STABLE3 mkdir ../squid-2.5.STABLE4 find . -depth -print | cpio -pdv ../squid-1.1.11 cd ../squid-1.1.11 patch -p1 < /tmp/squid-2.5.STABLE3-STABLE4.diff or alternatively cp -rl squid-2.5.STABLE3 squid-2.5.STABLE4 cd squid-2.5.STABLE4 zcat /tmp/squid-2.5.STABLE3-STABLE4.diff.gz | patch -p1 After the patch has been applied, you must rebuild Squid from the very beginning, i.e.: make distclean ./configure ... make make install

If your , for example. The configure script can take numerous options. The most useful is /usr/local/squid/. To change the default, you could do: % cd squid-x.y.z % ./configure --prefix=/some/other/directory/squid

Type % ./configure --help to see all available options. You will need to specify some of these options to enable or disable certain features. Some options which are used often include: --prefix=PREFIX install architecture-independent files in PREFIX [/usr/local/squid] --enable-dlmalloc[=LIB] Compile & use the malloc package by Doug Lea --enable-gnuregex Compile GNUregex --enable-splaytree Use SPLAY trees to store ACL lists --enable-xmalloc-debug Do some simple malloc debugging --enable-xmalloc-debug-trace Detailed trace of memory allocations --enable-xmalloc-statistics Show malloc statistics in status page --enable-carp Enable CARP support --enable-async-io Do ASYNC disk I/O using threads --enable-icmp Enable ICMP pinging --enable-delay-pools Enable delay pools to limit bandwith usage --enable-mem-gen-trace Do trace of memory stuff --enable-useragent-log Enable logging of User-Agent header --enable-kill-parent-hack Kill parent on shutdown --enable-snmp Enable SNMP monitoring --enable-cachemgr-hostname[=hostname] Make cachemgr.cgi default to this host --enable-arp-acl Enable use of ARP ACL lists (ether address) --enable-htpc Enable HTCP protocol --enable-forw-via-db Enable Forw/Via database --enable-cache-digests Use Cache Digests see http://www.squid-cache.org/Doc/FAQ/FAQ-16.html --enable-err-language=lang Select language for Error pages (see errors dir) undefined reference to __inet_ntoa

by and .

Probably you've recently installed bind 8.x. There is a mismatch between the header files and DNS library that Squid has found. There are a couple of things you can try.

First, try adding src/Makefile. If If that doesn't seem to work, edit your arpa/inet.h file and comment out the following: #define inet_addr __inet_addr #define inet_aton __inet_aton #define inet_lnaof __inet_lnaof #define inet_makeaddr __inet_makeaddr #define inet_neta __inet_neta #define inet_netof __inet_netof #define inet_network __inet_network #define inet_net_ntop __inet_net_ntop #define inet_net_pton __inet_net_pton #define inet_ntoa __inet_ntoa #define inet_pton __inet_pton #define inet_ntop __inet_ntop #define inet_nsap_addr __inet_nsap_addr #define inet_nsap_ntoa __inet_nsap_ntoa How can I get true DNS TTL info into Squid's IP cache?

If you have source for BIND, you can modify it as indicated in the diff below. It causes the global variable _dns_ttl_ to be set with the TTL of the most recent lookup. Then, when you compile Squid, the configure script will look for the _dns_ttl_ symbol in libresolv.a. If found, dnsserver will return the TTL value for every lookup.

This hack was contributed by . diff -ru bind-4.9.4-orig/res/gethnamaddr.c bind-4.9.4/res/gethnamaddr.c --- bind-4.9.4-orig/res/gethnamaddr.c Mon Aug 5 02:31:35 1996 +++ bind-4.9.4/res/gethnamaddr.c Tue Aug 27 15:33:11 1996 @@ -133,6 +133,7 @@ } align; extern int h_errno; +int _dns_ttl_; #ifdef DEBUG static void @@ -223,6 +224,7 @@ host.h_addr_list = h_addr_ptrs; haveanswer = 0; had_error = 0; + _dns_ttl_ = -1; while (ancount-- > 0 && cp < eom && !had_error) { n = dn_expand(answer->buf, eom, cp, bp, buflen); if ((n < 0) || !(*name_ok)(bp)) { @@ -232,8 +234,11 @@ cp += n; /* name */ type = _getshort(cp); cp += INT16SZ; /* type */ - class = _getshort(cp); - cp += INT16SZ + INT32SZ; /* class, TTL */ + class = _getshort(cp); + cp += INT16SZ; /* class */ + if (qtype == T_A && type == T_A) + _dns_ttl_ = _getlong(cp); + cp += INT32SZ; /* TTL */ n = _getshort(cp); cp += INT16SZ; /* len */ if (class != C_IN) {

And here is a patch for BIND-8: *** src/lib/irs/dns_ho.c.orig Tue May 26 21:55:51 1998 --- src/lib/irs/dns_ho.c Tue May 26 21:59:57 1998 *************** *** 87,92 **** --- 87,93 ---- #endif extern int h_errno; + int _dns_ttl_; /* Definitions. */ *************** *** 395,400 **** --- 396,402 ---- pvt->host.h_addr_list = pvt->h_addr_ptrs; haveanswer = 0; had_error = 0; + _dns_ttl_ = -1; while (ancount-- > 0 && cp < eom && !had_error) { n = dn_expand(ansbuf, eom, cp, bp, buflen); if ((n < 0) || !(*name_ok)(bp)) { *************** *** 404,411 **** cp += n; /* name */ type = ns_get16(cp); cp += INT16SZ; /* type */ ! class = ns_get16(cp); ! cp += INT16SZ + INT32SZ; /* class, TTL */ n = ns_get16(cp); cp += INT16SZ; /* len */ if (class != C_IN) { --- 406,416 ---- cp += n; /* name */ type = ns_get16(cp); cp += INT16SZ; /* type */ ! class = _getshort(cp); ! cp += INT16SZ; /* class */ ! if (qtype == T_A && type == T_A) ! _dns_ttl_ = _getlong(cp); ! cp += INT32SZ; /* TTL */ n = ns_get16(cp); cp += INT16SZ; /* len */ if (class != C_IN) { My platform is BSD/OS or BSDI and I can't compile Squid

cache_cf.c: In function `parseConfigFile': cache_cf.c:1353: yacc stack overflow before `token' ...

You may need to upgrade your gcc installation to a more recent version. Check your gcc version with gcc -v If it is earlier than 2.7.2, you might consider upgrading. Problems compiling The following error occurs on Solaris systems using gcc when the Solaris C compiler is not installed: /usr/bin/rm -f libmiscutil.a /usr/bin/false r libmiscutil.a rfc1123.o rfc1738.o util.o ... make[1]: *** [libmiscutil.a] Error 255 make[1]: Leaving directory `/tmp/squid-1.1.11/lib' make: *** [all] Error 1 Note on the second line the /usr/bin/false. This is supposed to be a path to the To fix this you either need to: Add /usr/ccs/bin to your PATH. This is where the Install the . This package includes programs such as I have problems compiling Squid on Platform Foo.

Please check the on which Squid is known to compile. Your problem might be listed there together with a solution. If it isn't listed there, mail us what you are trying, your Squid version, and the problems you encounter. I see a lot warnings while compiling Squid.

Warnings are usually not a big concern, and can be common with software designed to operate on multiple platforms. If you feel like fixing compile-time warnings, please do so and send us the patches. Building Squid on OS/2


In order in compile squid, you need to have a reasonable facsimile of a Unix system installed. This includes I made a few modifications to the pristine EMX 0.9d install. added defines for changed all occurrences of time_t to signed long instead of unsigned long hacked ld.exe to search for both xxxx.a and libxxxx.a to produce the correct filename when using the -Zexe option

You will need to run scripts/convert.configure.to.os2 (in the Squid source distribution) to modify the configure script so that it can search for the various programs.

Next, you need to set a few environment variables (see EMX docs for meaning): export EMXOPT="-h256 -c" export LDFLAGS="-Zexe -Zbin -s"

Now you are ready to configure squid: ./configure

Compile everything: make

and finally, install: make install

This will by default, install into /usr/local/squid. If you wish to install somewhere else, see the Now, don't forget to set EMXOPT before running squid each time. I recommend using the -Y and -N options. Installing and Running Squid How big of a system do I need to run Squid?

There are no hard-and-fast rules. The most important resource for Squid is physical memory. Your processor does not need to be ultra-fast. Your disk system will be the major bottleneck, so fast disks are important for high-volume caches. Do not use IDE disks if you can help it.

In late 1998, if you are buying a new machine for a cache, I would recommend the following configuration: 300 MHz Pentium II CPU 512 MB RAM Five 9 GB UW-SCSI disks Your system disk, and logfile disk can probably be IDE without losing any cache performance.

Also, see by Martin Hamilton This is a very nice page summarizing system configurations people are using for large Squid caches. How do I install Squid?

After , you can install it with this simple command: % make install If you have enabled the then you will also want to type % su # make install-pinger

After installing, you will want to edit and customize the /usr/local/squid/etc/squid.conf.

Also, a QUICKSTART guide has been included with the source distribution. Please see the directory where you unpacked the source archive. What does the The Do you have a Yes, after you How do I start Squid?

First you need to make your Squid configuration. The Squid configuration can be found in /usr/local/squid/etc/squid.conf and by default includes documentation on all directives.

In the Suqid distribution there is a small QUICKSTART guide indicating which directives you need to look closer at and why. At a absolute minimum you need to change the http_access configuration to allow access from your clients.

To verify your configuration file you can use the -k parse option % /usr/local/squid/sbin/squid -k parse If this outputs any errors then these are syntax errors or other fatal misconfigurations and needs to be corrected before you continue. If it is silent and immediately gives back the command promt then your squid.conf is syntactically correct and could be understood by Squid.

After you've finished editing the configuration file, you can start Squid for the first time. The procedure depends a little bit on which version you are using.

First, you must create the swap directories. Do this by running Squid with the -z option: % /usr/local/squid/sbin/squid -z

NOTE: If you run Squid as root then you may need to first create /usr/local/squid/var/logs and your cache_dir directories and assign ownership of these to the cache_effective_user configured in your squid.conf.

Once the creation of the cache directories completes, you can start Squid and try it out. Probably the best thing to do is run it from your terminal and watch the debugging output. Use this command: % /usr/local/squid/sbin/squid -NCd1 If everything is working okay, you will see the line: Ready to serve requests. If you want to run squid in the background, as a daemon process, just leave off all options: % /usr/local/squid/sbin/squid

NOTE: depending on which http_port you select you may need to start squid as root (http_port <1024).

NOTE: In Squid-2.4 and earlier Squid was installed in bin by default, not sbin. How do I start Squid automatically when the system boots?

Squid-2 has a restart feature built in. This greatly simplifies starting Squid and means that you don't need to use /usr/local/squid/sbin/squid

Squid will automatically background itself and then spawn a child process. In your Sep 23 23:55:58 kitty squid[14616]: Squid Parent: child process 14617 started That means that process ID 14563 is the parent process which monitors the child process (pid 14617). The child process is the one that does all of the work. The parent process just waits for the child process to exit. If the child process exits unexpectedly, the parent will automatically start another child process. In that case, Sep 23 23:56:02 kitty squid[14616]: Squid Parent: child process 14617 exited with status 1 Sep 23 23:56:05 kitty squid[14616]: Squid Parent: child process 14619 started

If there is some problem, and Squid can not start, the parent process will give up after a while. Your Sep 23 23:56:12 kitty squid[14616]: Exiting due to repeated, frequent failures When this happens you should check your When you look at a process ( 24353 ?? Ss 0:00.00 /usr/local/squid/bin/squid 24354 ?? R 0:03.39 (squid) (squid) The first is the parent process, and the child process is the one called ``(squid)''. Note that if you accidentally kill the parent process, the child process will not notice.

If you want to run Squid from your termainal and prevent it from backgrounding and spawning a child process, use the /usr/local/squid/bin/squid -N From inittab

On systems which have an /etc/inittab file (Digital Unix, Solaris, IRIX, HP-UX, Linux), you can add a line like this: sq:3:respawn:/usr/local/squid/sbin/squid.sh < /dev/null >> /tmp/squid.log 2>&1 We recommend using a #!/bin/sh C=/usr/local/squid PATH=/usr/bin:$C/bin TZ=PST8PDT export PATH TZ # User to notify on restarts notify="root" # Squid command line options opts="" cd $C umask 022 sleep 10 while [ -f /var/run/nosquid ]; do sleep 1 done /usr/bin/tail -20 $C/logs/cache.log \ | Mail -s "Squid restart on `hostname` at `date`" $notify exec bin/squid -N $opts From rc.local

On BSD-ish systems, you will need to start Squid from the ``rc'' files, usually /etc/rc.local. For example: if [ -f /usr/local/squid/sbin/squid ]; then echo -n ' Squid' /usr/local/squid/sbin/squid fi From init.d

Squid ships with a init.d type startup script in contrib/squid.rc which works on most init.d type systems. Or you can write your own using any normal init.d script found in your system as template and add the start/stop fragments shown below.

Start: /usr/local/squid/sbin/squid

Stop: /usr/local/squid/sbin/squid -k shutdown n=120 while /usr/local/squid/sbin/squid -k check && [ $n -gt 120 ]; do sleep 1 echo -n . n=`expr $n - 1` done How do I tell if Squid is running?

You can use the % squidclient http://www.netscape.com/ > test

There are other command-line HTTP client programs available as well. Two that you may find useful are and .

Another way is to use Squid itself to see if it can signal a running Squid process: % squid -k check And then check the shell's exit status variable.

Also, check the log files, most importantly the These are the command line options for How do I see how Squid works?

Check the Install and use the . Can Squid benefit from SMP systems?

Squid is a single process application and can not make use of SMP. If you want to make Squid benefit from a SMP system you will need to run multiple instances of Squid and find a way to distribute your users on the different Squid instances just as if you had multiple Squid boxes.

Having two CPUs is indeed nice for running other CPU intensive tasks on the same server as the proxy, such as if you have a lot of logs and need to run various statistics collections during peak hours.

The authentication and group helpers barely use any CPU and does not benefit from dual-CPU configuration. Is it okay to use separate drives and RAID on Squid?

RAID1 is fine, and so are separate drives.

RAID0 (striping) with Squid only gives you the drawback that if you loose one of the drives the whole stripe set is lost. There is no benefit in performance as Squid already distributes the load on the drives quite nicely.

Squid is the worst case application for RAID5, whether hardware or software, and will absolutely kill the performance of a RAID5. Once the cache has been filled Squid uses a lot of small random writes which the worst case workload for RAID5, effectively reducing write speed to only little more than that of one single drive.

Generally seek time is what you want to optimize for Squid, or more precisely the total amount of seeks/s your system can sustain. Choosing the right RAID solution generally decreases the amount of seeks/s your system can sustain significantly. Configuration issues How do I join a cache hierarchy?

To place your cache in a hierarchy, use the For example, the following # squid.conf - On the host: childcache.example.com # # Format is: hostname type http_port udp_port # cache_peer parentcache.example.com parent 3128 3130 cache_peer childcache2.example.com sibling 3128 3130 cache_peer childcache3.example.com sibling 3128 3130 The # squid.conf - On the host: sv.cache.nlanr.net # # Format is: hostname type http_port udp_port # cache_peer electraglide.geog.unsw.edu.au parent 3128 3130 cache_peer cache1.nzgate.net.nz parent 3128 3130 cache_peer pb.cache.nlanr.net parent 3128 3130 cache_peer it.cache.nlanr.net parent 3128 3130 cache_peer sd.cache.nlanr.net parent 3128 3130 cache_peer uc.cache.nlanr.net sibling 3128 3130 cache_peer bo.cache.nlanr.net sibling 3128 3130 cache_peer_domain electraglide.geog.unsw.edu.au .au cache_peer_domain cache1.nzgate.net.nz .au .aq .fj .nz cache_peer_domain pb.cache.nlanr.net .uk .de .fr .no .se .it cache_peer_domain it.cache.nlanr.net .uk .de .fr .no .se .it cache_peer_domain sd.cache.nlanr.net .mx .za .mu .zm The configuration above indicates that the cache will use How do I join NLANR's cache hierarchy?

We have a simple set of the NLANR cache hierarchy. Why should I want to join NLANR's cache hierarchy?

The NLANR hierarchy can provide you with an initial source for parent or sibling caches. Joining the NLANR global cache system will frequently improve the performance of your caching service. How do I register my cache with NLANR's registration service?

Just enable these options in your cache_announce 24 announce_to sd.cache.nlanr.net:3131 How do I find other caches close to me and arrange parent/child/sibling relationships with them?

Visit the NLANR cache to discover other caches near you. Keep in mind that just because a cache is registered in the database My cache registration is not appearing in the Tracker database.

Your site will not be listed if your cache IP address does not have a DNS PTR record. If we can't map the IP address back to a domain name, it will be listed as ``Unknown.'' The registration messages are sent with UDP. We may not be receiving your announcement message due to firewalls which block UDP, or dropped packets due to congestion. What is the httpd-accelerator mode?

This entry has been moved to . How do I configure Squid to work behind a firewall?

Note: The information here is current for version 2.2.

If you are behind a firewall then you can't make direct connections to the outside world, so you You can use the acl INSIDE dstdomain .mydomain.com always_direct allow INSIDE never_direct allow all

You could also specify internal servers by IP address acl INSIDE_IP dst always_direct allow INSIDE_IP never_direct allow all Note, however that when you use IP addresses, Squid must perform a DNS lookup to convert URL hostnames to an address. Your internal DNS servers may not be able to lookup external domains.

If you use cache_peer xyz.mydomain.com parent 3128 0 default How do I configure Squid forward all requests to another proxy?

Note: The information here is current for version 2.2.

First, you need to give Squid a parent cache. Second, you need to tell Squid it can not connect directly to origin servers. This is done with three configuration file lines: cache_peer parentcache.foo.com parent 3128 0 no-query default acl all src never_direct allow all Note, with this configuration, if the parent cache fails or becomes unreachable, then every request will result in an error message.

In case you want to be able to use direct connections when all the parents go down you should use a different approach: cache_peer parentcache.foo.com parent 3128 0 no-query prefer_direct off The default behaviour of Squid in the absence of positive ICP, HTCP, etc replies is to connect to the origin server instead of using parents. The prefer_direct off directive tells Squid to try parents first. I have The It's very important that there are enough My First, find out if you have enough Another factor which affects the How can I easily change the default HTTP port?

Before you run the configure script, simply set the setenv CACHE_HTTP_PORT 8080 ./configure make make install Is it possible to control how big each With Squid-1.1 it is NOT possible. Each What Most people have a disk partition dedicated to the Squid cache. You don't want to use the entire partition size. You have to leave some extra room. Currently, Squid is not very tolerant of running out of disk space.

Lets say you have a 9GB disk. Remember that disk manufacturers lie about the space available. A so-called 9GB disk usually results in about 8.5GB of raw, usable space. First, put a filesystem on it, and mount it. Then check the ``available space'' with your Next, I suggest taking off another 10% or so for Squid overheads, and a "safe buffer." Squid normally puts its cache_dir ... 7000 16 256

Its better to start out conservative. After the cache becomes full, look at the disk usage. If you think there is plenty of unused space, then increase the If you're getting ``disk full'' write errors, then you definately need to decrease your cache size. I'm adding a new With Squid-1.1, yes, you will lose your cache. This is because version 1.1 uses a simplistic algorithm to distribute files between cache directories.

With Squid-2, you will not lose your existing cache. You can add and delete Squid and Several people on both the . The most elegant way in my opinion is to run an internal Squid caching proxyserver which handles client requests and let this server forward it's requests to the http-gw running on the firewall. Cache hits won't need to be handled by the firewall.

In this example Squid runs on the same server as the http-gw, Squid uses 8000 and http-gw uses 8080 (web). The local domain is Firewall configuration:

Either run http-gw as a daemon from the /etc/rc.d/rc.local (Linux Slackware): exec /usr/local/fwtk/http-gw -daemon 8080 or run it from inetd like this: web stream tcp nowait.100 root /usr/local/fwtk/http-gw http-gw I increased the watermark to 100 because a lot of people run into problems with the default value.

Make sure you have at least the following line in /usr/local/etc/netperm-table: http-gw: hosts You could add the IP-address of your own workstation to this rule and make sure the http-gw by itself works, like: http-gw: hosts Squid configuration:

The following settings are important: http_port 8000 icp_port 0 cache_peer localhost.home.nl parent 8080 0 default acl HOME dstdomain .home.nl alwayws_direct allow HOME never_direct allow all This tells Squid to use the parent for all domains other than 872739961.631 1566 ERR_CLIENT_ABORT/304 83 GET http://www.squid-cache.org/ - DEFAULT_PARENT/localhost.home.nl - 872739962.976 1266 TCP_CLIENT_REFRESH/304 88 GET http://www.nlanr.net/Images/cache_now.gif - DEFAULT_PARENT/localhost.home.nl - 872739963.007 1299 ERR_CLIENT_ABORT/304 83 GET http://www.squid-cache.org/Icons/squidnow.gif - DEFAULT_PARENT/localhost.home.nl - 872739963.061 1354 TCP_CLIENT_REFRESH/304 83 GET http://www.squid-cache.org/Icons/Squidlogo2.gif - DEFAULT_PARENT/localhost.home.nl

http-gw entries in syslog: Aug 28 02:46:00 memo http-gw[2052]: permit host=localhost/ use of gateway (V2.0beta) Aug 28 02:46:00 memo http-gw[2052]: log host=localhost/ protocol=HTTP cmd=dir dest=www.squid-cache.org path=/ Aug 28 02:46:01 memo http-gw[2052]: exit host=localhost/ cmds=1 in=0 out=0 user=unauth duration=1 Aug 28 02:46:01 memo http-gw[2053]: permit host=localhost/ use of gateway (V2.0beta) Aug 28 02:46:01 memo http-gw[2053]: log host=localhost/ protocol=HTTP cmd=get dest=www.squid-cache.org path=/Icons/Squidlogo2.gif Aug 28 02:46:01 memo http-gw[2054]: permit host=localhost/ use of gateway (V2.0beta) Aug 28 02:46:01 memo http-gw[2054]: log host=localhost/ protocol=HTTP cmd=get dest=www.squid-cache.org path=/Icons/squidnow.gif Aug 28 02:46:01 memo http-gw[2055]: permit host=localhost/ use of gateway (V2.0beta) Aug 28 02:46:01 memo http-gw[2055]: log host=localhost/ protocol=HTTP cmd=get dest=www.nlanr.net path=/Images/cache_now.gif Aug 28 02:46:02 memo http-gw[2055]: exit host=localhost/ cmds=1 in=0 out=0 user=unauth duration=1 Aug 28 02:46:03 memo http-gw[2053]: exit host=localhost/ cmds=1 in=0 out=0 user=unauth duration=2 Aug 28 02:46:04 memo http-gw[2054]: exit host=localhost/ cmds=1 in=0 out=0 user=unauth duration=3

To summarize:

Advantages: http-gw allows you to selectively block ActiveX and Java, and it's primary design goal is security. The firewall doesn't need to run large applications like Squid. The internal Squid-server still gives you the benefit of caching.

Disadvantages: The internal Squid proxyserver can't (and shouldn't) work with other parent or neighbor caches. Initial requests are slower because these go through http-gw, http-gw also does reverse lookups. Run a nameserver on the firewall or use an internal nameserver. -- What is ``HTTP_X_FORWARDED_FOR''? Why does squid provide it to WWW servers, and how can I stop it?

When a proxy-cache is used, a server does not see the connection coming from the originating client. Many people like to implement access controls based on the client address. To accommodate these people, Squid adds its own request header called "X-Forwarded-For" which looks like this: X-Forwarded-For:, unknown, Entries are always IP addresses, or the word We must note that access controls based on this header are extremely weak and simple to fake. Anyone may hand-enter a request with any IP address whatsoever. This is perhaps the reason why client IP addresses have been omitted from the HTTP/1.1 specification.

Because of the weakness of this header support for access controls based on X-Forwarder-For is not yet available in any officially released version of squid. However, unofficial patches are available from the Squid development project and may be integrated into later versions of Squid once a suitable trust model have been developed. Can Squid anonymize HTTP requests?

Yes it can, however the way of doing it has changed from earlier versions of squid. As of squid-2.2 a more customisable method has been introduced. Please follow the instructions for the version of squid that you are using. As a default, no anonymizing is done.

If you choose to use the anonymizer you might wish to investigate the forwarded_for option to prevent the client address being disclosed. Failure to turn off the forwarded_for option will reduce the effectiveness of the anonymizer. Finally if you filter the User-Agent header using the fake_user_agent option can prevent some user problems as some sites require the User-Agent header. Squid 2.2

With the introduction of squid 2.2 the anonoymizer has become more customisable. It now allows specification of exactly which headers will be allowed to pass. This is further extended in Squid-2.5 to allow headers to be anonymized conditionally.

For details see the documentation of the http_header_access and header_replace directives in squid.conf.default.

References: Can I make Squid go direct for some sites?

Sure, just use the For example, if you want Squid to connect directly to acl hotmail dstdomain .hotmail.com always_direct allow hotmail Can I make Squid proxy only, without caching anything?

Sure, there are few things you can do.

You can use the acl all src 0/0 no_cache deny all

With Squid-2.4 and later you can use the ``null'' storage module to avoid having a cache directory: cache_dir null /tmp

Note: a null cache_dir does not disable caching, but it does save you from creating a cache structure if you have disabled caching with no_cache.

Note: the directory (e.g., /tmp) must exist so that squid can chdir to it, unless you also use the To configure Squid for the ``null'' storage module, specify it on the ./configure --enable-storeio=ufs,null ... Can I prevent users from downloading large files?

You can set the global If the HTTP response coming from the server has a Some responses don't have Note that ``creative'' user-agents will still be able to download really large files through the cache using HTTP/1.1 range requests. Communication between browsers and Squid

Most web browsers available today support proxying and are easily configured to use a Squid server as a proxy. Some browsers support advanced features such as lists of domains or URL patterns that shouldn't be fetched through the proxy, or JavaScript automatic proxy configuration. Netscape manual configuration

Select Here is a of the Netscape Navigator manual proxy configuration screen.

Netscape automatic configuration

Netscape Navigator's proxy configuration can be automated with JavaScript (for Navigator versions 2.0 or higher). Select Here is a of the Netscape Navigator automatic proxy configuration screen. You may also wish to consult Netscape's documentation for the Navigator

Here is a sample auto configuration JavaScript from Oskar Pearson: //We (www.is.co.za) run a central cache for our customers that they //access through a firewall - thus if they want to connect to their intranet //system (or anything in their domain at all) they have to connect //directly - hence all the "fiddling" to see if they are trying to connect //to their local domain. //Replace each occurrence of company.com with your domain name //and if you have some kind of intranet system, make sure //that you put it's name in place of "internal" below. //We also assume that your cache is called "cache.company.com", and //that it runs on port 8080. Change it down at the bottom. //(C) Oskar Pearson and the Internet Solution (http://www.is.co.za) function FindProxyForURL(url, host) { //If they have only specified a hostname, go directly. if (isPlainHostName(host)) return "DIRECT"; //These connect directly if the machine they are trying to //connect to starts with "intranet" - ie http://intranet //Connect directly if it is intranet.* //If you have another machine that you want them to //access directly, replace "internal*" with that //machine's name if (shExpMatch( host, "intranet*")|| shExpMatch(host, "internal*")) return "DIRECT"; //Connect directly to our domains (NB for Important News) if (dnsDomainIs( host,"company.com")|| //If you have another domain that you wish to connect to //directly, put it in here dnsDomainIs(host,"sistercompany.com")) return "DIRECT"; //So the error message "no such host" will appear through the //normal Netscape box - less support queries :) if (!isResolvable(host)) return "DIRECT"; //We only cache http, ftp and gopher if (url.substring(0, 5) == "http:" || url.substring(0, 4) == "ftp:"|| url.substring(0, 7) == "gopher:") //Change the ":8080" to the port that your cache //runs on, and "cache.company.com" to the machine that //you run the cache on return "PROXY cache.company.com:8080; DIRECT"; //We don't cache WAIS if (url.substring(0, 5) == "wais:") return "DIRECT"; else return "DIRECT"; } Lynx and Mosaic configuration

For Mosaic and Lynx, you can set environment variables before starting the application. For example (assuming csh or tcsh):

% setenv http_proxy http://mycache.example.com:3128/ % setenv gopher_proxy http://mycache.example.com:3128/ % setenv ftp_proxy http://mycache.example.com:3128/

For Lynx you can also edit the http_proxy:http://mycache.example.com:3128/ ftp_proxy:http://mycache.example.com:3128/ gopher_proxy:http://mycache.example.com:3128/ Redundant Proxy Auto-Configuration

There's one nasty side-effect to using auto-proxy scripts: if you start the web browser it will try and load the auto-proxy-script.

If your script isn't available either because the web server hosting the script is down or your workstation can't reach the web server (e.g. because you're working off-line with your notebook and just want to read a previously saved HTML-file) you'll get different errors depending on the browser you use.

The Netscape browser will just return an error after a timeout (after that it tries to find the site 'www.proxy.com' if the script you use is called 'proxy.pac').

The Microsoft Internet Explorer on the other hand won't even start, no window displays, only after about 1 minute it'll display a window asking you to go on with/without proxy configuration.

The point is that your workstations always need to locate the proxy-script. I created some extra redundancy by hosting the script on two web servers (actually Apache web servers on the proxy servers themselves) and adding the following records to my primary nameserver: proxy CNAME proxy1 CNAME proxy2 The clients just refer to 'http://proxy/proxy.pac'. This script looks like this: function FindProxyForURL(url,host) { // Hostname without domainname or host within our own domain? // Try them directly: // http://www.domain.com actually lives before the firewall, so // make an exception: if ((isPlainHostName(host)||dnsDomainIs( host,".domain.com")) && !localHostOrDomainIs(host, "www.domain.com")) return "DIRECT"; // First try proxy1 then proxy2. One server mostly caches '.com' // to make sure both servers are not // caching the same data in the normal situation. The other // server caches the other domains normally. // If one of 'm is down the client will try the other server. else if (shExpMatch(host, "*.com")) return "PROXY proxy1.domain.com:8080; PROXY proxy2.domain.com:8081; DIRECT"; return "PROXY proxy2.domain.com:8081; PROXY proxy1.domain.com:8080; DIRECT"; }

I made sure every client domain has the appropriate 'proxy' entry. The clients are automatically configured with two nameservers using DHCP. -- Proxy Auto-Configuration with URL Hashing

The contains a lot of good information about hash-based proxy auto-configuration scripts. With these you can distribute the load between a number of caching proxies. Microsoft Internet Explorer configuration

Select Here is a of the Internet Explorer proxy configuration screen.

Microsoft is also starting to support Netscape-style JavaScript automated proxy configuration. As of now, only MSIE version 3.0a for Windows 3.1 and Windows NT 3.51 supports this feature (i.e., as of version 3.01 build 1225 for Windows 95 and NT 4.0, the feature was not included).

If you have a version of MSIE that does have this feature, elect Netmanage Internet Chameleon WebSurfer configuration

Netmanage WebSurfer supports manual proxy configuration and exclusion lists for hosts or domains that should not be fetched via proxy (this information is current as of WebSurfer 5.0). Select Take a look at this if the instructions confused you.

On the same configuration window, you'll find a button to bring up the exclusion list dialog box, which will let you enter some hosts or domains that you don't want fetched via proxy. It should be self-explanatory, but you might look at this just for fun anyway. Opera 2.12 proxy configuration

Select Notes: Opera 2.12 doesn't support gopher on its own, but requires a proxy; therefore Squid's gopher proxying can extend the utility of your Opera immensely. Unfortunately, Opera 2.12 chokes on some HTTP requests, for example . At the moment I think it has something to do with cookies. If you have trouble with a site, try disabling the HTTP proxying by unchecking that protocol in the -- How do I tell Squid to use a specific username for FTP urls?

Insert your username in the host part of the URL, for example: ftp://joecool@ftp.foo.org/ Squid should then prompt you for your account password. Alternatively, you can specify both your username and password in the URL itself: ftp://joecool:secret@ftp.foo.org/ However, we certainly do not recommend this, as it could be very easy for someone to see or grab your password. Configuring Browsers for WPAD


You may like to start by reading the that describes WPAD.

After reading the 8 steps below, if you don't understand any of the terms or methods mentioned, you probably shouldn't be doing this. Implementing wpad requires you to web server installations and modifications. squid proxy server (or others) installation etc. Domain Name System maintenance etc. Please don't bombard the squid list with web server or dns questions. See your system administrator, or do some more research on those topics.

This is not a recommendation for any product or version. As far as I know IE5 is the only browser out now implementing wpad. I think wpad is an excellent feature that will return several hours of life per month. Hopefully, all browser clients will implement it as well. But it will take years for all the older browsers to fade away though.

I have only focused on the domain name method, to the exclusion of the DHCP method. I think the dns method might be easier for most people. I don't currently, and may never, fully understand wpad and IE5, but this method worked for me. It But if you'd rather just have a go ... Create a standard . The sample provided there is more than adequate to get you going. No doubt all the other load balancing and backup scripts will be fine also. Store the resultant file in the document root directory of a handy web server as notes that you should be able to use an HTTP redirect if you want to store the wpad.dat file somewhere else. You can probably even redirect Redirect /wpad.dat http://racoon.riga.lv/proxy.pac If you do nothing more, a url like http://www.your.domain.name/wpad.dat should bring up the script text in your browser window. Insert the following entry into your web server application/x-ns-proxy-autoconfig dat And then restart your web server, for new mime type to work. Assuming Internet Explorer 5, under http://www.your.domain.name/wpad.dat Test that that all works as per your script and network. There's no point continuing until this works ... Create/install/implement a DNS record so that wpad.your.domain.name resolves to the host above where you have a functioning auto config script running. You should now be able to use http://wpad.your.domain.name/wpad.dat as the Auto Config Script location in step 5 above. And finally, go back to the setup screen detailed in 5 above, and choose nothing but the One final question might be 'Which domain name does the client (IE5) use for the wpad... lookup?' It uses the hostname from the control panel setting. It starts the search by adding the hostname "WPAD" to current fully-qualified domain name. For instance, a client in a.b.Microsoft.com would search for a WPAD server at wpad.a.b.microsoft.com. If it could not locate one, it would remove the bottom-most domain and try again; for instance, it would try wpad.b.microsoft.com next. IE 5 would stop searching when it found a WPAD server or reached the third-level domain, wpad.microsoft.com.

Anybody using these steps to install and test, please feel free to make notes, corrections or additions for improvements, and post back to the squid list...

There are probably many more tricks and tips which hopefully will be detailed here in the future. Things like Configuring Browsers for WPAD with DHCP

You can also use DHCP to configure browsers for WPAD. This technique allows you to set any URL as the PAC URL. For ISC DHCPD, enter a line like this in your option wpad code 252 = text; option wpad "http://www.example.com/proxy.pac";

Replace the hostname with the name or address of your own server.

Ilja Pavkovic notes that the DHCP mode does not work reliably with every version of Internet Explorer. The DNS name method to find wpad.dat is more reliable.

Another user adds that IE 6.01 seems to strip the last character from the URL. By adding a trailing newline, he is able to make it work with both IE 5.0 and 6.0:< option wpad "http://www.example.com/proxy.pac\n"; IE 5.0x crops trailing slashes from FTP URL's


There was a bug in the 5.0x releases of Internet Explorer in which IE cropped any trailing slash off an FTP URL. The URL showed up correctly in the browser's ``Address:'' field, however squid logs show that the trailing slash was being taken off.

An example of where this impacted squid if you had a setup where squid would go direct for FTP directory listings but forward a request to a parent for FTP file transfers. This was useful if your upstream proxy was an older version of Squid or another vendors software which displayed directory listings with broken icons and you wanted your own local version of squid to generate proper FTP directory listings instead. The workaround for this is to add a double slash to any directory listing in which the slash was important, or else upgrade to IE 5.5. (Or use Netscape) IE 6.0 SP1 fails when using authentication

When using authentication with Internet Explorer 6 SP1, you may encounter issues when you first launch Internet Explorer. The problem will show itself when you first authenticate, you will receive a "Page Cannot Be Displayed" error. However, if you click refresh, the page will be correctly displayed.

This only happens immediately after you authenticate.

This is not a Squid error or bug. Microsoft broke the Basic Authentication when they put out IE6 SP1.

There is a knowledgebase article () regarding this issue, which contains a link to a downloadable "hot fix." They do warn that this code is not "regression tested" but so far there have not been any reports of this breaking anything else. The problematic file is wininet.dll. Please note that this hotfix is included in the latest security update.

Lloyd Parkes notes that the article references another article, . He says that you must According to Joao Coutinho, this simple solution also corrects the problem: Go to Tools/Internet Go to Options/Advanced UNSELECT "Show friendly HTTP error messages" under Browsing.

Another possible workaround to these problems is to make the ERR_CACHE_ACCESS_DENIED larger than 1460 bytes. This should trigger IE to handle the authentication in a slightly different manner. Squid Log Files

The logs are a valuable source of information about Squid workloads and performance. The logs record not only access information, but also system configuration errors and resource consumption (eg, memory, disk space). There are several log file maintained by Squid. Some have to be explicitely activated during compile time, others can safely be deactivated during run-time.

There are a few basic points common to all log files. The time stamps logged into the log files are usually UTC seconds unless stated otherwise. The initial time stamp usually contains a millisecond extension. If you run your Squid from the The From the area of automatic log file analysis, the The user agent log file is only maintained, if you configured the compile time you pointed the

From the user agent log file you are able to find out about distributation of browsers of your clients. Using this option in conjunction with a loaded production squid might not be the best of all ideas. The The The print format for a store log entry (one line) consists of eleven space-separated columns, compare with the src/store_log.c: "%9d.%03d %-7s %08X %4d %9d %9d %9d %s %d/%d %s %s\n" The timestamp when the line was logged in UTC with a millisecond fraction. The action the object was sumitted to, compare with src/store_log.c: ).

The file number for the object storage file. Please note that the path to this file is calculated according to your A file number of The HTTP reply status code.

The value of the HTTP "Date: " reply header.

The value of the HTTP "Last-Modified: " reply header.

The value of the HTTP "Expires: " reply header. The HTTP "Content-Type" major value, or "unknown" if it cannot be determined. This column consists of two slash separated fields: The advertised content length from the HTTP "Content-Length: " reply header. The size actually read.

If the advertised (or expected) length is missing, it will be set to zero. If the advertised length is not zero, but not equal to the real length, the object will be realeased from the cache. The request method for the object, e.g.

The key to the object, usually the URL.

The timestamp format for the columns to are all expressed in UTC seconds. The actual values are parsed from the HTTP reply headers. An unparsable header is represented by a value of -1, and a missing header is represented by a value of -2.

The column usually contains just the URL of the object. Some objects though will never become public. Thus the key is said to include a unique integer number and the request method in addition to the URL. This logfile exists for Squid-1.0 only. The format is [date] URL peerstatus peerhost Most log file analysis program are based on the entries in The common log file format contains other information than the native log file, and less. The native format contains more information for the admin interested in cache evaluation. The is used by numerous HTTP servers. This format consists of the following seven fields: remotehost rfc931 authuser [date] "method URL" status bytes

It is parsable by a variety of tools. The common format contains different information than the native log file format. The HTTP version is logged, which is not logged in native log file format. The native format is different for different major versions of Squid. For Squid-1.0 it is: time elapsed remotehost code/status/peerstatus bytes method URL

For Squid-1.1, the information from the time elapsed remotehost code/status bytes method URL rfc931 peerstatus/peerhost type

For Squid-2 the columns stay the same, though the content within may change a little.

The native log file format logs more and different information than the common log file format: the request duration, some timeout information, the next upstream server address, and the content type. There exist tools, which convert one file format into the other. Please mind that even though the log formats share most information, both formats contain information which is not part of the other format, and thus this part of the information is lost when converting. Especially converting back and forth is not possible without loss. It is recommended though to use Squid's native log format due to its greater amount of information made available for later analysis. The print format line for native "%9d.%03d %6d %s %s/%03d %d %s %s %s %s%s/%s %s"

Therefore, an A Unix timestamp as UTC seconds with a millisecond resolution. You can convert Unix timestamps into something more human readable using this short perl script: #! /usr/bin/perl -p s/^\d+\.\d+/localtime $&/e; The elapsed time considers how many milliseconds the transaction busied the cache. It differs in interpretation between TCP and UDP:

For HTTP/1.0, this is basically the time between For persistent connections, this ought to be the time between scheduling the reply and finishing sending it. For ICP, this is the time between scheduling a reply and actually sending it.

Please note that the entries are logged The IP address of the requesting instance, the client IP address. The Also, the

This column is made up of two entries separated by a slash. This column encodes the transaction result: The cache result of the request contains information on the kind of request, how it was satisfied, or in what way it failed. Please refer to section for valid symbolic result codes.

Several codes from older versions are no longer available, were renamed, or split. Especially the for details on the codes no longer available in Squid-2.

The NOVM versions and Squid-2 also rely on the Unix buffer cache, thus you will see less The status part contains the HTTP result codes with some Squid specific extensions. Squid uses a subset of the RFC defined error codes for HTTP. Refer to section for details of the status codes recognized by a Squid-2. The size is the amount of data delivered to the client. Mind that this does not constitute the net object size, as headers are also counted. Also, failed requests may deliver an error page, the size of which is also logged here. The request method to obtain an object. Please refer to section for available methods. If you turned off This column contains the URL requested. Please note that the log file may contain whitespaces for the URI. The default configuration for The eigth column may contain the ident lookups for the requesting client. Since ident lookups have performance impact, the default configuration turns The hierarchy information consists of three items:

Any hierarchy tag may be prefixed with A code that explains how the request was handled, e.g. by forwarding it to a peer, or going straight to the source. Refer to section for details on hierarchy codes and removed hierarchy codes. The IP address or hostname where the request (if a miss) was forwarded. For requests sent to origin servers, this is the origin server's IP address. For requests sent to a neighbor cache, this is the neighbor's hostname. NOTE: older versions of Squid would put the origin server hostname here. The content type of the object as seen in the HTTP reply header. Please note that ICP exchanges usually don't have any content type, and thus are logged ``-''. Also, some weird replies have content types ``:'' or even empty ones.

There may be two more columns in the Squid result codes

The The following result codes were taken from a Squid-2, compare with the src/access_log.c: The client issued a "no-cache" pragma, or some analogous cache control command along with the request. Thus, the cache has to refetch the object. The client issued an IMS request for an object which was in the cache and fresh. The object was believed to be in the cache, but could not be accessed. During "-Y" startup, or during frequent failures, a cache in hit only mode will return either UDP_HIT or this code. Neighbours will thus only fetch hits.

The following codes are no longer available in Squid-2: . . used instead. . HTTP status codes

These are taken from and verified for Squid. Squid-2 uses almost all codes except 307 (Temporary Redirect), 416 (Request Range Not Satisfiable), and 417 (Expectation Failed). Extra codes include 0 for a result code being unavailable, and 600 to signal an invalid header, a proxy error. Also, some definitions were added as for (WebDAV). Yes, there are really two entries for status code 424, compare with src/enums.h: 000 Used mostly with UDP traffic. 100 Continue 101 Switching Protocols *102 Processing 200 OK 201 Created 202 Accepted 203 Non-Authoritative Information 204 No Content 205 Reset Content 206 Partial Content *207 Multi Status 300 Multiple Choices 301 Moved Permanently 302 Moved Temporarily 303 See Other 304 Not Modified 305 Use Proxy [307 Temporary Redirect] 400 Bad Request 401 Unauthorized 402 Payment Required 403 Forbidden 404 Not Found 405 Method Not Allowed 406 Not Acceptable 407 Proxy Authentication Required 408 Request Timeout 409 Conflict 410 Gone 411 Length Required 412 Precondition Failed 413 Request Entity Too Large 414 Request URI Too Large 415 Unsupported Media Type [416 Request Range Not Satisfiable] [417 Expectation Failed] *424 Locked *424 Failed Dependency *433 Unprocessable Entity 500 Internal Server Error 501 Not Implemented 502 Bad Gateway 503 Service Unavailable 504 Gateway Timeout 505 HTTP Version Not Supported *507 Insufficient Storage 600 Squid header parsing error Request methods

Squid recognizes several request methods as defined in . Newer versions of Squid (2.2.STABLE5 and above) also recognize ``HTTP Extensions for Distributed Authoring -- WEBDAV'' extensions. method defined cachabil. meaning --------- ---------- ---------- ------------------------------------------- GET HTTP/0.9 possibly object retrieval and simple searches. HEAD HTTP/1.0 possibly metadata retrieval. POST HTTP/1.0 CC or Exp. submit data (to a program). PUT HTTP/1.1 never upload data (e.g. to a file). DELETE HTTP/1.1 never remove resource (e.g. file). TRACE HTTP/1.1 never appl. layer trace of request route. OPTIONS HTTP/1.1 never request available comm. options. CONNECT HTTP/1.1r3 never tunnel SSL connection. ICP_QUERY Squid never used for ICP based exchanges. PURGE Squid never remove object from cache. PROPFIND rfc2518 ? retrieve properties of an object. PROPATCH rfc2518 ? change properties of an object. MKCOL rfc2518 never create a new collection. COPY rfc2518 never create a duplicate of src in dst. MOVE rfc2518 never atomically move src to dst. LOCK rfc2518 never lock an object against modifications. UNLOCK rfc2518 never unlock an object. Hierarchy Codes

The following hierarchy codes are used with Squid-2: src/peer_select.c:hier_strings[]. src/peer_select.c:hier_strings[].

Almost any of these may be preceded by 'TIMEOUT_' if the two-second (default) timeout occurs waiting for all ICP replies to arrive from neighbors, see also the The following hierarchy codes were removed from Squid-2: code meaning -------------------- ------------------------------------------------- PARENT_UDP_HIT_OBJ hit objects are not longer available. SIBLING_UDP_HIT_OBJ hit objects are not longer available. SSL_PARENT_MISS SSL can now be handled by squid. FIREWALL_IP_DIRECT No special logging for hosts inside the firewall. LOCAL_IP_DIRECT No special logging for local networks. cache/log (Squid-1.x)

This file has a rather unfortunate name. It also is often called the % squid -k shutdown This will disrupt service, but at least you will have your swap log back. Alternatively, you can tell squid to rotate its log files. This also causes a clean swap log to be written. % squid -k rotate

For Squid-1.1, there are six fields: swap.state (Squid-2.x)

In Squid-2, the swap log file is now called for information on the contents and format of that file.

If you remove % squid -k rotate Alternatively, you can tell Squid to shutdown and it will rewrite this file before it exits.

If you remove the By default the Which log files can I delete safely?

You should never delete If you accidentally delete The correct way to maintain your log files is with Squid's ``rotate'' feature. You should rotate your log files at least once per day. The current log files are closed and then renamed with numeric extensions (.0, .1, etc). If you want to, you can write your own scripts to archive or remove the old log files. If not, Squid will only keep up to If you set To rotate Squid's logs, simple use this command: squid -k rotate For example, use this cron entry to rotate the logs at midnight: 0 0 * * * /usr/local/squid/bin/squid -k rotate How can I disable Squid's log files?

To disable cache_access_log /dev/null

To disable cache_store_log none

To disable cache_log /dev/null

To disable cache_access_log none

To disable cache_store_log none

To disable cache_log /dev/null

My log files get very big!

You need to 0 0 * * * /usr/local/squid/bin/squid -k rotate I want to use another tool to maintain the log files.

If you set Managing log files

The preferred log file for analysis is the Depending on the disk space allocated for log file storage, it is recommended to set up a cron job which rotates the log files every 24, 12, or 8 hour. You will need to set your Before transport, the log files can be compressed during off-peak time. On the analysis host, the log file are concatinated into one file, so one file for 24 hours is the yield. Also note that with The EU project developed some to obey when handling and processing log files: Respect the privacy of your clients when publishing results. Keep logs unavailable unless anonymized. Most countries have laws on privacy protection, and some even on how long you are legally allowed to keep certain kinds of information. Rotate and process log files at least once a day. Even if you don't process the log files, they will grow quite large, see section . If you rely on processing the log files, reserve a large enough partition solely for log files. Keep the size in mind when processing. It might take longer to process log files than to generate them! Limit yourself to the numbers you are interested in. There is data beyond your dreams available in your log file, some quite obvious, others by combination of different views. Here are some examples for figures to watch: The hosts using your cache. The elapsed time for HTTP requests - this is the latency the user sees. Usually, you will want to make a distinction for HITs and MISSes and overall times. Also, medians are preferred over averages. The requests handled per interval (e.g. second, minute or hour). Why do I get ERR_NO_CLIENTS_BIG_OBJ messages so often?

This message means that the requested object was in ``Delete Behind'' mode and the user aborted the transfer. An object will go into ``Delete Behind'' mode if It is larger than It is being fetched from a neighbor which has the What does ERR_LIFETIME_EXP mean?

This means that a timeout occurred while the object was being transferred. Most likely the retrieval of this object was very slow (or it stalled before finishing) and the user aborted the request. However, depending on your settings for Retrieving ``lost'' files from the cache

I've been asked to retrieve an object which was accidentally destroyed at the source for recovery. So, how do I figure out where the things are so I can copy them out and strip off the headers?

The following method applies only to the Squid-1.1 versions:

Use grep to find the named object (Url) in the file. The first field in this file is an integer Then, find the file perl fileno-to-pathname.pl [-c squid.conf] file numbers are read on stdin, and pathnames are printed on stdout. Can I use Sort of. You can use cached.

Cached responses are logged with the SWAPOUT tag. Uncached responses are logged with the RELEASE tag.

However, your analysis must also consider that when a cached response is removed from the cache (for example due to cache replacement) it is also logged in Operational issues How do I see system level Squid statistics?

The Squid distribution includes a CGI utility called How can I find the biggest objects in my cache?

sort -r -n +4 -5 access.log | awk '{print $5, $7}' | head -25 I want to restart Squid with a clean cache

Note: The information here is current for version 2.2.

First of all, you must stop Squid of course. You can use the command: % squid -k shutdown

The fastest way to restart with an entirely clean cache is to over write the % echo "" > /cache1/swap.state Repeat that for every Another way, which takes longer, is to have squid recreate all the % cd /cache1 % mkdir JUNK % mv ?? swap.state* JUNK % rm -rf JUNK & Repeat this for your other % squid -z How can I proxy/cache Real Audio?

by , and

Point the RealPlayer at your Squid server's HTTP port (e.g. 3128). Using the Preferences->Transport tab, select The RealPlayer (and RealPlayer Plus) manual states: Use HTTP Only Select this option if you are behind a firewall and cannot receive data through TCP. All data will be streamed through HTTP. Note: You may not be able to receive some content if you select this option.

Again, from the documentation: RealPlayer 4.0 identifies itself to the firewall when making a request for content to a RealServer. The following string is attached to any URL that the Player requests using HTTP GET: /SmpDsBhgRl Thus, to identify an HTTP GET request from the RealPlayer, look for: http://[^/]+/SmpDsBhgRl The Player can also be identified by the mime type in a POST to the RealServer. The RealPlayer POST has the following mime type: "application/x-pncmd" Note that the first request is a POST, and the second has a '?' in the URL, so standard Squid configurations would treat it as non-cachable. It also looks rather ``magic.''

HTTP is an alternative delivery mechanism introduced with version 3 players, and it allows a reasonable approximation to ``streaming'' data - that is playing it as you receive it.

It isn't available in the general case: only if someone has made the realaudio file available via an HTTP server, or they're using a version 4 server, they've switched it on, and you're using a version 4 client. If someone has made the file available via their HTTP server, then it'll be cachable. Otherwise, it won't be (as far as we can tell.)

The more common RealAudio link connects via their own Some confusion arises because there is also a configuration option to use an HTTP proxy (such as Squid) with the Realaudio/RealVideo players. This is because the players can fetch the ``How can I purge an object from my cache?

Squid does not allow you to purge objects unless it is configured with access controls in acl PURGE method PURGE acl localhost src http_access allow PURGE localhost http_access deny PURGE The above only allows purge requests which come from the local host and denies all other purge requests.

To purge an object, you can use the squidclient -m PURGE http://www.miscreant.com/ If the purge was successful, you will see a ``200 OK'' response: HTTP/1.0 200 OK Date: Thu, 17 Jul 1997 16:03:32 GMT Server: Squid/1.1.14 If the object was not found in the cache, you will see a ``404 Not Found'' response: HTTP/1.0 404 Not Found Date: Thu, 17 Jul 1997 16:03:22 GMT Server: Squid/1.1.14 Using ICMP to Measure the Network

As of version 1.1.9, Squid is able to utilize ICMP Round-Trip-Time (RTT) measurements to select the optimal location to forward a cache miss. Previously, cache misses would be forwarded to the parent cache which returned the first ICP reply message. These were logged with FIRST_PARENT_MISS in the access.log file. Now we can select the parent which is closest (RTT-wise) to the origin server. Supporting ICMP in your Squid cache

It is more important that your parent caches enable the ICMP features. If you are acting as a parent, then you may want to enable ICMP on your cache. Also, if your cache makes RTT measurements, it will fetch objects directly if your cache is closer than any of the parents.

If you want your Squid cache to measure RTT's to origin servers, Squid must be compiled with the USE_ICMP option. This is easily accomplished by uncommenting "-DUSE_ICMP=1" in src/Makefile and/or src/Makefile.in.

An external program called % make install % su # make install-pinger There are three configuration file options for tuning the measurement database on your cache. Another option, Utilizing your parents database

Your parent caches can be asked to include the RTT measurements in their ICP replies. To do this, you must enable query_icmp on This causes a flag to be set in your outgoing ICP queries.

If your parent caches return ICMP RTT measurements then the eighth column of your access.log will have lines similar to: CLOSEST_PARENT_MISS/it.cache.nlanr.net In this case, it means that CLOSEST_DIRECT/www.sample.com Inspecting the database

The measurement database can be viewed from the cachemgr by selecting "Network Probe Database." Hostnames are aggregated into /24 networks. All measurements made are averaged over time. Measurements are made to specific hosts, taken from the URLs of HTTP requests. The recv and sent fields are the number of ICMP packets sent and received. At this time they are only informational.

A typical database entry looks something like this: Network recv/sent RTT Hops Hostnames 20/ 21 82.3 6.0 www.jisedu.org www.dozo.com bo.cache.nlanr.net 42.0 7.0 uc.cache.nlanr.net 48.0 10.0 pb.cache.nlanr.net 55.0 10.0 it.cache.nlanr.net 185.0 13.0 This means we have sent 21 pings to both www.jisedu.org and www.dozo.com. The average RTT is 82.3 milliseconds. The next four lines show the measured values from our parent caches. Since Why are so few requests logged as TCP_IMS_MISS?

When Squid receives an If the request is not forwarded, Squid replies to the IMS request according to the object in its cache. If the modification times are the same, then Squid returns TCP_IMS_HIT. If the modification times are different, then Squid returns TCP_IMS_MISS. In most cases, the cached object will not have changed, so the result is TCP_IMS_HIT. Squid will only return TCP_IMS_MISS if some other client causes a newer version of the object to be pulled into the cache. How can I make Squid NOT cache some servers or URLs?

In Squid-2, you use the acl Local dst no_cache deny Local

This example makes all URL's with '.html' uncachable: acl HTML url_regex .html$ no_cache deny HTML

This example makes a specific URL uncachable: acl XYZZY url_regex ^http://www.i.suck.com/foo.html$ no_cache deny XYZZY

This example caches nothing between the hours of 8AM to 11AM: acl Morning time 08:00-11:00 no_cache deny Morning

In Squid-1.1, whether or not an object gets cached is controlled by the cache_stoplist my.domain.com Specifying uncachable objects by IP address is harder. The includes a patch called How can I delete and recreate a cache directory?

Deleting an existing cache directory is not too difficult. Unfortunately, you can't simply change squid.conf and then reconfigure. You can't stop using a cache_dir while Squid is running. Also note that Squid requires at least one cache_dir to run. Edit your If you don't have any cache_dir lines in your squid.conf, then Squid was using the default. You'll need to add a new cache_dir line because Squid will continue to use the default otherwise. You can add a small, temporary directory, fo example: /usr/local/squid/cachetmp .... If you add a new cache_dir you have to run squid -z to initialize that directory. Remeber that you can not delete a cache directory from a running Squid process; you can not simply reconfigure squid. You must shutdown Squid: squid -k shutdown Once Squid exits, you may immediately start it up again. Since you deleted the old cache_dir from squid.conf, Squid won't try to access that directory. If you use the RunCache script, Squid should start up again automatically. Now Squid is no longer using the cache directory that you removed from the config file. You can verify this by checking "Store Directory" information with the cache manager. From the command line, type: squidclient mgr:storedir Now that Squid is not using the cache directory, you can

The procedure is similar to recreate the directory. Edit Initialize the new directory by running % squid -z NOTE: it is safe to run this even if Squid is already running. Reconfigure Squid squid -k reconfigure Unlike deleting, you can add new cache directories while Squid is already running. Why can't I run Squid as root?

by Dave J Woolley

If someone were to discover a buffer overrun bug in Squid and it runs as a user other than root, they can only corrupt the files writeable to that user, but if it runs a root, they can take over the whole machine. This applies to all programs that don't absolutely need root status, not just squid. Can you tell me a good way to upgrade Squid with minimal downtime?

Here is a technique that was described by .

Start a second Squid server on an unused HTTP port (say 4128). This instance of Squid probably doesn't need a large disk cache. When this second server has finished reloading the disk store, swap the Can Squid listen on more than one HTTP port?

Note: The information here is current for version 2.3.

Yes, you can specify multiple With version 2.3 and later you can specify IP addresses and port numbers together (see the squid.conf comments). Can I make origin servers see the client's IP address when going through Squid?

Normally you cannot. Most TCP/IP stacks do not allow applications to create sockets with the local endpoint assigned to a foreign IP address. However, some folks have some that allow exactly that.

In this situation, you must ensure that all HTTP packets destined for the client IP addresses are routed to the Squid box. If the packets take another path, the real clients will send TCP resets to the origin servers, thereby breaking the connections. Memory

Squid uses a lot of memory for performance reasons. It takes much, much longer to read something from disk than it does to read directly from memory.

A small amount of metadata for each cached object is kept in memory. This is the Squid-1.1 also uses a lot of memory to store in-transit objects. This version stores incoming objects only in memory, until the transfer is complete. At that point it decides whether or not to store the object on disk. This means that when users download large files, your memory usage will increase significantly. The squid.conf parameter Other uses of memory by Squid include: Disk buffers for reading and writing Network I/O buffers IP Cache contents FQDN Cache contents Netdb ICMP measurement database Per-request state information, including full request and reply headers Miscellaneous statistics collection. ``Hot objects'' which are kept entirely in memory. How can I tell how much memory my Squid process is using?

One way is to simply look at wessels ˜ 236% ps -axuhm USER PID %CPU %MEM VSZ RSS TT STAT STARTED TIME COMMAND squid 9631 4.6 26.4 141204 137852 ?? S 10:13PM 78:22.80 squid -NCYs For SYSV-ish, you probably want to use the A nicer way to check the memory usage is with a program called last pid: 20128; load averages: 0.06, 0.12, 0.11 14:10:58 46 processes: 1 running, 45 sleeping CPU states: % user, % nice, % system, % interrupt, % idle Mem: 187M Active, 1884K Inact, 45M Wired, 268M Cache, 8351K Buf, 1296K Free Swap: 1024M Total, 256K Used, 1024M Free PID USERNAME PRI NICE SIZE RES STATE TIME WCPU CPU COMMAND 9631 squid 2 0 138M 135M select 78:45 3.93% 3.93% squid

Finally, you can ask the Squid process to report its own memory usage. This is available on the Cache Manager Resource usage for squid: Maximum Resident Size: 137892 KB Memory usage for squid via mstats(): Total space in arena: 140144 KB Total free: 8153 KB 6%

If your RSS (Resident Set Size) value is much lower than your process size, then your cache performance is most likely suffering due to . My Squid process grows without bounds.

You might just have your '' entry below.

When a process continually grows in size, without levelling off or slowing down, it often indicates a memory leak. A memory leak is when some chunk of memory is used, but not free'd when it is done being used.

Memory leaks are a real problem for programs (like Squid) which do all of their processing within a single process. Historically, Squid has had real memory leak problems. But as the software has matured, we believe almost all of Squid's memory leaks have been eliminated, and new ones are least easy to identify.

Memory leaks may also be present in your system's libraries, such as . I set The See also . How do I analyze memory usage from the cache manger output?

Note: This information is specific to Squid-1.1 versions

Look at your Memory usage for squid via mallinfo(): Total space in arena: 94687 KB Ordinary blocks: 32019 KB 210034 blks Small blocks: 44364 KB 569500 blks Holding blocks: 0 KB 5695 blks Free Small blocks: 6650 KB Free Ordinary blocks: 11652 KB Total in use: 76384 KB 81% Total free: 18302 KB 19% Meta Data: StoreEntry 246043 x 64 bytes = 15377 KB IPCacheEntry 971 x 88 bytes = 83 KB Hash link 2 x 24 bytes = 0 KB URL strings = 11422 KB Pool MemObject structures 514 x 144 bytes = 72 KB ( 70 free) Pool for Request structur 516 x 4380 bytes = 2207 KB ( 2121 free) Pool for in-memory object 6200 x 4096 bytes = 24800 KB ( 22888 free) Pool for disk I/O 242 x 8192 bytes = 1936 KB ( 1888 free) Miscellaneous = 2600 KB total Accounted = 58499 KB

First note that Of that 94M, 81% (76M) is actually being used at the moment. The rest has been freed, or pre-allocated by Of the 76M in use, we can account for 58.5M (76%). There are some calls to The The pool sizes are specified by Pool for disk I/O is hardcoded at 200. For If you need to lower your process size, we recommend lowering the max object sizes in the 'http', 'ftp' and 'gopher' config lines. You may also want to lower The ``Total memory accounted'' value is less than the size of my Squid process.

We are not able to account for Also, note that the xmalloc: Unable to allocate 4096 bytes!


Messages like "FATAL: xcalloc: Unable to allocate 4096 blocks of 1 bytes!" appear when Squid can't allocate more memory, and on most operating systems (inclusive BSD) there are only two possible reasons: The machine is out of swap The process' maximum data segment size has been reached The first case is detected using the normal swap monitoring tools available on the platform ( To tell if it is the second case, first rule out the first case and then monitor the size of the Squid process. If it dies at a certain size with plenty of swap left then the max data segment size is reached without no doubts.

The data segment size can be limited by two factors: Kernel imposed maximum, which no user can go above The size set with ulimit, which the user can control.

When squid starts it sets data and file ulimit's to the hard level. If you manually tune ulimit before starting Squid make sure that you set the hard limit and not only the soft limit (the default operation of ulimit is to only change the soft limit). root is allowed to raise the soft limit above the hard limit.

This command prints the hard limits: ulimit -aH

This command sets the data size to unlimited: ulimit -HSd unlimited BSD/OS


The default kernel limit on BSD/OS for datasize is 64MB (at least on 3.0 which I'm using).

Recompile a kernel with larger datasize settings: maxusers 128 # Support for large inpcb hash tables, e.g. busy WEB servers. options INET_SERVER # support for large routing tables, e.g. gated with full Internet routing: options "KMEMSIZE=\(16*1024*1024\)" options "DFLDSIZ=\(128*1024*1024\)" options "DFLSSIZ=\(8*1024*1024\)" options "SOMAXCONN=128" options "MAXDSIZ=\(256*1024*1024\)" See /usr/share/doc/bsdi/config.n for more info.

In /etc/login.conf I have this: default:\ :path=/bin /usr/bin /usr/contrib/bin:\ :datasize-cur=256M:\ :openfiles-cur=1024:\ :openfiles-max=1024:\ :maxproc-cur=1024:\ :stacksize-cur=64M:\ :radius-challenge-styles=activ,crypto,skey,snk,token:\ :tc=auth-bsdi-defaults:\ :tc=auth-ftp-bsdi-defaults: # # Settings used by /etc/rc and root # This must be set properly for daemons started as root by inetd as well. # Be sure reset these values back to system defaults in the default class! # daemon:\ :path=/bin /usr/bin /sbin /usr/sbin:\ :widepasswords:\ :tc=default: # :datasize-cur=128M:\ # :openfiles-cur=256:\ # :maxproc-cur=256:\

This should give enough space for a 256MB squid process. FreeBSD (2.2.X)

by Duane Wessels

The procedure is almost identical to that for BSD/OS above. Increase the open filedescriptor limit in /sys/conf/param.c: int maxfiles = 4096; int maxfilesperproc = 1024; Increase the maximum and default data segment size in your kernel config file, e.g. /sys/conf/i386/CONFIG: options "MAXDSIZ=(512*1024*1024)" options "DFLDSIZ=(128*1024*1024)" We also found it necessary to increase the number of mbuf clusters: options "NMBCLUSTERS=10240" And, if you have more than 256 MB of physical memory, you probably have to disable BOUNCE_BUFFERS (whatever that is), so comment out this line: #options BOUNCE_BUFFERS #include support for DMA bounce buffers Also, update limits in /etc/login.conf: # Settings used by /etc/rc # daemon:\ :coredumpsize=infinity:\ :datasize=infinity:\ :maxproc=256:\ :maxproc-cur@:\ :memoryuse-cur=64M:\ :memorylocked-cur=64M:\ :openfiles=4096:\ :openfiles-cur@:\ :stacksize=64M:\ :tc=default: And don't forget to run ``cap_mkdb /etc/login.conf'' after editing that file. OSF, Digital Unix


To increase the data size for Digital UNIX, edit the file /etc/sysconfigtab and add the entry... proc: per-proc-data-size=1073741824 Or, with csh, use the limit command, such as > limit datasize 1024M

Editing /etc/sysconfigtab requires a reboot, but the limit command doesn't. fork: (12) Cannot allocate memory

When Squid is reconfigured (SIGHUP) or the logs are rotated (SIGUSR1), some of the helper processes (dnsserver) must be killed and restarted. If your system does not have enough virtual memory, the Squid process may not be able to fork to start the new helper processes. This is due to the UNIX way of starting child processes using the fork() system call which temporary duplicates the whole Squid process, and when rapidly starting many child processes such as on "squid -k rotate" the memory usage can temporarily grow to many times the normal memory usage due to several temporary copies of the whole process.

The best way to fix this is to increase your virtual memory by adding swap space. Normally your system uses raw disk partitions for swap space, but most operating systems also support swapping on regular files (Digital Unix excepted). See your system manual pages for What can I do to reduce Squid's memory usage?

If your cache performance is suffering because of memory limitations, you might consider buying more memory. But if that is not an option, There are a number of things to try: Try a . Reduce the Turn the Reduce the Reduce the If you are using Squid-1.1.x, try the ``NOVM'' version. Using an alternate

Many users have found improved performance and memory utilization when linking Squid with an external malloc library. We recommend either GNU malloc, or dlmalloc. Using GNU malloc

To make Squid use GNU malloc follow these simple steps: Download the GNU malloc source, available from one of . Compile GNU malloc % gzip -dc malloc.tar.gz | tar xf - % cd malloc % vi Makefile # edit as needed % make Copy libmalloc.a to your system's library directory and be sure to name it % su # cp malloc.a /usr/lib/libgnumalloc.a (Optional) Copy the GNU malloc.h to your system's include directory and be sure to name it # cp malloc.h /usr/include/gnumalloc.h Reconfigure and recompile Squid % make realclean % ./configure ... % make % make install Note, In later distributions, 'realclean' has been changed to 'distclean'. As the configure script runs, watch its output. You should find that it locates libgnumalloc.a and optionally gnumalloc.h. dlmalloc

has been written by . According to Doug: This is not the fastest, most space-conserving, most portable, or most tunable malloc ever written. However it is among the fastest while also being among the most space-conserving, portable and tunable.

dlmalloc is included with the % ./configure --enable-dlmalloc ... How much memory do I need in my Squid server?

As a rule of thumb on Squid uses approximately 10 MB of RAM per GB of the total of all cache_dirs (more on 64 bit servers such as Alpha), plus your cache_mem setting and about an additional 10-20MB. It is recommended to have at least twice this amount of physical RAM available on your Squid server. For a more detailed discussion on Squid's memory usage see the sections above.

The recommended extra RAM besides what is used by Squid is used by the operating system to improve disk I/O performance and by other applications or services running on the server. This will be true even of a server which runs Squid as the only tcp service, since there is a minimum level of memory needed for process management, logging, and other OS level routines.

If you have a low memory server, and a large disk, then you will not necessarily be able to use all the disk space, since as the cache fills the memory available will be insufficient, forcing Squid to swap out memory and affecting performance. A very large cache_dir total and insufficient physical RAM + Swap could cause Squid to stop functioning completely. The solution for larger caches is to get more physical RAM; allocating more to Squid via cache_mem will not help. The Cache Manager

by What is the cache manager?

The cache manager (How do you set it up?

That depends on which web server you're using. Below you will find instructions for configuring the CERN and Apache servers to permit After you edit the server configuration files, you will probably need to either restart your web server or or send it a When you're done configuring your web server, you'll connect to the cache manager with a web browser, using a URL such as: http://www.example.com/Squid/cgi-bin/cachemgr.cgi/ Cache manager configuration for CERN httpd 3.0

First, you should ensure that only specified workstations can access the cache manager. That is done in your CERN Protection MGR-PROT { Mask @(workstation.example.com) } Wildcards are acceptable, IP addresses are acceptable, and others can be added with a comma-separated list of IP addresses. There are many more ways of protection. Your server documentation has details.

You also need to add: Protect /Squid/* MGR-PROT Exec /Squid/cgi-bin/*.cgi /usr/local/squid/bin/*.cgi This marks the script as executable to those in Cache manager configuration for Apache

First, make sure the cgi-bin directory you're using is listed with a ScriptAlias /Squid/cgi-bin/ /usr/local/squid/cgi-bin/ It's probably a Next, you should ensure that only specified workstations can access the cache manager. That is done in your Apache <Location /Squid/cgi-bin/cachemgr.cgi> order allow,deny allow from workstation.example.com &etago;Location> You can have more than one allow line, and you can allow domains or networks.

Alternately, <Location /Squid/cgi-bin/cachemgr.cgi> AuthUserFile /path/to/password/file AuthGroupFile /dev/null AuthName User/Password Required AuthType Basic require user cachemanager &etago;Location> Consult the Apache documentation for information on using Cache manager configuration for Roxen 2.0 and later

by Francesco ``kinkie'' Chemolli

Notice: this is This is what's required to start up a fresh Virtual Server, only serving the cache manager. If you already have some Virtual Server you wish to use to host the Cache Manager, just add a new CGI support module to it.

Create a new virtual server, and set it to host http://www.example.com/. Add to it at least the following modules: Content Types CGI scripting support

In the CGI-bin path: set to /Squid/cgi-bin/ Handle *.cgi: set to Run user scripts as owner: set to Search path: set to the directory containing the cachemgr.cgi file

In section allow ip= where is the IP address for workstation.example.com

Save the configuration, and you're done. Cache manager ACLs in The default cache manager access configuration in acl manager proto cache_object acl localhost src acl all src With the following rules: http_access deny manager !localhost http_access allow all

The first ACL is the most important as the cache manager program interrogates squid using a special telnet mycache.example.com 3128 GET cache_object://mycache.example.com/info HTTP/1.0

The default ACLs say that if the request is for a In fact, only allowing localhost access means that on the initial acl manager proto cache_object acl localhost src acl example src acl all src Where http_access allow manager localhost http_access allow manager example http_access deny manager http_access allow all If you're using miss_access allow manager

The default ACLs assume that your web server is on the same machine as Always be sure to send a Why does it say I need a password and a URL?

If you ``drop'' the list box, and browse it, you will see that the password is only required to shutdown the cache, and the URL is required to refresh an object (i.e., retrieve it from its original source again) Otherwise these fields can be left blank: a password is not required to obtain access to the informational aspects of I want to shutdown the cache remotely. What's the password?

See the How do I make the cache host default to When you run % ./configure --enable-cachemgr-hostname=`hostname` ...

Note, if you do this after you already installed Squid before, you need to make sure % cd src % rm cachemgr.o cachemgr.cgi % make cachemgr.cgi

Then copy What's the difference between Squid TCP connections and Squid UDP connections?

Browsers and caches use TCP connections to retrieve web objects from web servers or caches. UDP connections are used when another cache using you as a sibling or parent wants to find out if you have an object in your cache that it's looking for. The UDP connections are ICP queries. It says the storage expiration will happen in 1970!

Don't worry. The default (and sensible) behavior of What do the Meta Data entries mean?

Basically just like the Info about objects currently in memory, (eg, in the process of being transferred). Information about each request as it happens. Space for object data as it is retrieved.

If In the utilization section, what is In the utilization section, why is the Transfer KB/sec column always zero?

This column contains gross estimations of data transfer rates averaged over the entire time the cache has been running. These numbers are unreliable and mostly useless. In the utilization section, what is the The number of objects of that type in the cache right now. In the utilization section, what is the Max/Current/Min KB?

These refer to the size all the objects of this type have grown to/currently are/shrunk to. What is the I/O section about?

These are histograms on the number of bytes read from the network per What is the this will download to your browser a list of every URL in the cache and statistics about it. It can be very, very large. You probably don't need this information anyway. What is the What does Average Round Trip Time. This is how long on average after an ICP ping is sent that a reply is received. In the IP cache section, what's the difference between a hit, a negative hit and a miss?

A HIT means that the document was found in the cache. A MISS, that it wasn't found in the cache. A negative hit means that it was found in the cache, but it doesn't exist. What do the IP cache contents mean anyway?

The hostname is the name that was requested to be resolved.

For the The The The rest of the line lists all the IP addresses that have been associated with that IP cache entry.

What is the fqdncache and how is it different from the ipcache?

IPCache contains data for the Hostname to IP-Number mapping, and FQDNCache does it the other way round. For example: Hostname Flags lstref TTL N [IP-Number] gorn.cc.fh-lippe.de C 0 21581 1 lagrange.uni-paderborn.de C 6 21594 1 www.altavista.digital.com C 10 21299 4 ... 2/ftp.symantec.com DL 1583 -772855 0 Flags: C --> Cached D --> Dispatched N --> Negative Cached L --> Locked lstref: Time since last use TTL: Time-To-Live until information expires N: Count of addresses

IP-Number Flags TTL N Hostname C -45570 1 andele.cs.tu-berlin.de C -58133 1 komet.teuto.de N -73747 0 Flags: C --> Cached D --> Dispatched N --> Negative Cached L --> Locked TTL: Time-To-Live until information expires N: Count of names What does ``Page faults with physical i/o: 4897'' mean?

This question was asked on the by

You get a ``page fault'' when your OS tries to access something in memory which is actually swapped to disk. The term ``page fault'' while correct at the kernel and CPU level, is a bit deceptive to a user, as there's no actual error - this is a normal feature of operation.

Also, this doesn't necessarily mean your squid is swapping by that much. Most operating systems also implement paging for executables, so that only sections of the executable which are actually used are read from disk into memory. Also, whenever squid needs more memory, the fact that the memory was allocated will show up in the page faults.

However, if the number of faults is unusually high, and getting bigger, this could mean that squid is swapping. Another way to verify this is using a program called ``vmstat'' which is found on most UNIX platforms. If you run this as ``vmstat 5'' this will update a display every 5 seconds. This can tell you if the system as a whole is swapping a lot (see your local man page for vmstat for more information).

It is very bad for squid to swap, as every single request will be blocked until the requested data is swapped in. It is better to tweak the by

There's two different operations at work, Paging and swapping. Paging is when individual pages are shuffled (either discarded or swapped to/from disk), while ``swapping'' Needless to say, swapping a process is a pretty drastic event, and usually only reserved for when there's a memory crunch and paging out cannot free enough memory quickly enough. Also, there's some variation on how swapping is implemented in OS's. Some don't do it at all or do a hybrid of paging and swapping instead.

As you say, paging out doesn't necessarily involve disk IO, eg: text (code) pages are read-only and can simply be discarded if they are not used (and reloaded if/when needed). Data pages are also discarded if unmodified, and paged out if there's been any changes. Allocated memory (malloc) is always saved to disk since there's no executable file to recover the data from. mmap() memory is variable.. If it's backed from a file, it uses the same rules as the data segment of a file - ie: either discarded if unmodified or paged out.

There's also ``demand zeroing'' of pages as well that cause faults.. If you malloc memory and it calls brk()/sbrk() to allocate new pages, the chances are that you are allocated demand zero pages. Ie: the pages are not ``really'' attached to your process yet, but when you access them for the first time, the page fault causes the page to be connected to the process address space and zeroed - this saves unnecessary zeroing of pages that are allocated but never used.

The ``page faults with physical IO'' comes from the OS via getrusage(). It's highly OS dependent on what it means. Generally, it means that the process accessed a page that was not present in memory (for whatever reason) and there was disk access to fetch it. Many OS's load executables by demand paging as well, so the act of starting squid implicitly causes page faults with disk IO - however, many (but not all) OS's use ``read ahead'' and ``prefault'' heuristics to streamline the loading. Some OS's maintain ``intent queues'' so that pages can be selected as pageout candidates ahead of time. When (say) squid touches a freshly allocated demand zero page and one is needed, the OS can page out one of the candidates on the spot, causing a 'fault with physical IO' with demand zeroing of allocated memory which doesn't happen on many other OS's. (The other OS's generally put the process to sleep while the pageout daemon finds a page for it).

The meaning of ``swapping'' varies. On FreeBSD for example, swapping out is implemented as unlocking upages, kernel stack, PTD etc for aggressive pageout with the process. The only thing left of the process in memory is the 'struct proc'. The FreeBSD paging system is highly adaptive and can resort to paging in a way that is equivalent to the traditional swapping style operation (ie: entire process). FreeBSD also tries stealing pages from active processes in order to make space for disk cache. I suspect this is why setting 'memory_pools off' on the non-NOVM squids on FreeBSD is reported to work better - the VM/buffer system could be competing with squid to cache the same pages. It's a pity that squid cannot use mmap() to do file IO on the 4K chunks in it's memory pool (I can see that this is not a simple thing to do though, but that won't stop me wishing. :-).


The comments so far have been about what paging/swapping figures mean in a ``traditional'' context, but it's worth bearing in mind that on some systems (Sun's Solaris 2, at least), the virtual memory and filesystem handling are unified and what a user process sees as reading or writing a file, the system simply sees as paging something in from disk or a page being updated so it needs to be paged out. (I suppose you could view it as similar to the operating system memory-mapping the files behind-the-scenes.)

The effect of this is that on Solaris 2, paging figures will also include file I/O. Or rather, the figures from vmstat certainly appear to include file I/O, and I presume (but can't quickly test) that figures such as those quoted by Squid will also include file I/O.

To confirm the above (which represents an impression from what I've read and observed, rather than 100% certain facts...), using an otherwise idle Sun Ultra 1 system system I just tried using cat (small, shouldn't need to page) to copy (a) one file to another, (b) a file to /dev/null, (c) /dev/zero to a file, and (d) /dev/zero to /dev/null (interrupting the last two with control-C after a while!), while watching with vmstat. 300-600 page-ins or page-outs per second when reading or writing a file (rather than a device), essentially zero in other cases (and when not cat-ing).

So ... beware assuming that all systems are similar and that paging figures represent *only* program code and data being shuffled to/from disk - they may also include the work in reading/writing all those files you were accessing... Ok, so what is unusually high?

You'll probably want to compare the number of page faults to the number of HTTP requests. If this ratio is close to, or exceeding 1, then Squid is paging too much. What does the IGNORED field mean in the 'cache server list'?

This refers to ICP replies which Squid ignored, for one of these reasons: The URL in the reply could not be found in the cache at all. The URL in the reply was already being fetched. Probably this ICP reply arrived too late. The URL in the reply did not have a MemObject associated with it. Either the request is already finished, or the user aborted before the ICP arrived. The reply came from a multicast-responder, but the Source-Echo replies from known neighbors are ignored. ICP_OP_DENIED replies are ignored after the first 100. Access Controls

Squid's access control scheme is relatively comprehensive and difficult for some people to understand. There are two different components: ACL elements

Note: The information here is current for version 2.5.

Squid knows about the following types of ACL elements:


Not all of the ACL elements can be used with all types of access lists (described below). For example, The The SNMP ACL element and access list require the --enable-snmp configure option.

Some ACL elements can cause processing delays. For example, use of Each ACL element is assigned a unique You can't give the same name to two different types of ACL elements. It will generate a syntax error.

You can put different values for the same ACL name on different lines. Squid combines them into one list. Access Lists

There are a number of different access lists:


An access list An access list consists of one or more access list rules.

Access list rules are checked in the order they are written. List searching terminates as soon as one of the rules is a match.

If a rule has multiple ACL elements, it uses AND logic. In other words, To summarise the acl logics can be described as: http_access allow|deny acl AND acl AND ... OR http_access allow|deny acl AND acl AND ... OR ...

If none of the rules are matched, then the default action is the acl all src 0/0 http_access deny all How do I allow my clients to use the cache?

Define an ACL that corresponds to your client's IP addresses. For example: acl myclients src Next, allow those clients in the http_access allow myclients how do I configure Squid not to cache a specific server?

acl someserver dstdomain .someserver.com no_cache deny someserver How do I implement an ACL ban list?

As an example, we will assume that you would like to prevent users from accessing cooking recipes.

One way to implement this would be to deny access to any URLs that contain the words ``cooking'' or ``recipe.'' You would use these configuration lines: acl Cooking1 url_regex cooking acl Recipe1 url_regex recipe http_access deny Cooking1 http_access deny Recipe1 http_access allow all The Another way is to deny access to specific servers which are known to hold recipes. For example: acl Cooking2 dstdomain www.gourmet-chef.com http_access deny Cooking2 http_access allow all The How do I block specific users or groups from accessing my cache? Ident

You can use to allow specific users access to your cache. This requires that an process runs on the user's machine(s). In your ident_lookup_access allow all acl friends ident kim lisa frank joe http_access allow friends http_access deny all Proxy Authentication

Another option is to use proxy-authentication. In this scheme, you assign usernames and passwords to individuals. When they first use the proxy they are asked to authenticate themselves by entering their username and password.

In Squid v2 this authentication is hanled via external processes. For information on how to configure this, please see . Do you have a CGI program which lets users change their own proxy passwords?

has adapted the Apache's . Is there a way to do ident lookups only for a certain host and compare the result with a userlist in squid.conf?

You can use the requests.

Additionally, if you use a However, Squid does not wait for the lookup to complete unless the ACL rules require it. Consider this configuration: acl host1 src acl host2 src acl pals ident kim lisa frank joe http_access allow host1 http_access allow host2 pals Requests coming from will be allowed immediately because there are no user requirements for that host. However, requests from will be allowed only after the ident lookup completes, and if the username is in the set kim, lisa, frank, or joe. Common Mistakes And/Or logic

You've probably noticed (and been frustrated by) the fact that you cannot combine access controls with terms like ``and'' or ``or.'' These operations are already built in to the access control scheme in a fundamental way which you must understand. All elements of an . All elements of an . e.g.

For example, the following access control configuration will never work: acl ME src acl YOU src http_access allow ME YOU In order for the request to be allowed, it must match the ``ME'' acl AND the ``YOU'' acl. This is impossible because any IP address could only match one or the other. This should instead be rewritten as: acl ME src acl YOU src http_access allow ME http_access allow YOU Or, alternatively, this would also work: acl US src http_access allow US allow/deny mixups

I have read through my squid.conf numerous times, spoken to my neighbors, read the FAQ and Squid Docs and cannot for the life of me work out why the following will not work.

I can successfully access cachemgr.cgi from our web server machine here, but I would like to use MRTG to monitor various aspects of our proxy. When I try to use 'squidclient' or GET cache_object from the machine the proxy is running on, I always get access denied. acl manager proto cache_object acl localhost src acl server src acl all src acl ourhosts src http_access deny manager !localhost !server http_access allow ourhosts http_access deny all

The intent here is to allow cache manager requests from the http_access deny manager !localhost !server

The problem here is that for allowable requests, this access rule is not matched. For example, if the source IP address is To implement the desired policy correctly, the access rules should be rewritten as http_access allow manager localhost http_access allow manager server http_access deny manager http_access allow ourhosts http_access deny all If you're using miss_access allow manager

You may be concerned that the having five access rules instead of three may have an impact on the cache performance. In our experience this is not the case. Squid is able to handle a moderate amount of access control checking without degrading overall performance. You may like to verify that for yourself, however. Differences between For the I set up my access controls, but they don't work! why?

If ACLs are giving you problems and you don't know why they aren't working, you can use this tip to debug them.

In squid.conf enable debugging for section 33 at level 2. For example: debug_options ALL,1 33,2 Then restart or reconfigure squid.

From now on, your If this does not give you sufficient information to nail down the problem you can also enable detailed debug information on ACL processing debug_options ALL,1 33,2 28,9 Then restart or reconfigure squid as above.

From now on, your See also Proxy-authentication and neighbor caches

The problem... [ Parents ] / \ / \ [ Proxy A ] --- [ Proxy B ] | | USER

Proxy A sends and ICP query to Proxy B about an object, Proxy B replies with an ICP_HIT. Proxy A forwards the HTTP request to Proxy B, but does not pass on the authentication details, therefore the HTTP GET from Proxy A fails.

Only ONE proxy cache in a chain is allowed to ``use'' the Proxy-Authentication request header. Once the header is used, it must not be passed on to other proxies.

Therefore, you must allow the neighbor caches to request from each other without proxy authentication. This is simply accomplished by listing the neighbor ACL's first in the list of acl proxy-A src acl proxy-B src acl user_passwords proxy_auth /tmp/user_passwds http_access allow proxy-A http_access allow proxy-B http_access allow user_passwords http_access deny all Is there an easy way of banning all Destination addresses except one?

acl GOOD dst acl BAD dst http_access allow GOOD http_access deny BAD Does anyone have a ban list of porn sites and such?

Snerpa, an ISP in Iceland operates a DNS-database of IP-addresses of blacklisted sites containing porn, violence, etc. which is utilized using a small perl-script redirector. Information on this on the webpage. The redirector folks provide a blacklist. Bill Stearns maintains the of known spammers. By blocking the spammer web sites in squid, users can no longer use up bandwidth downloading spam images and html. Even more importantly, they can no longer send out requests for things like scripts and gifs that have a unique identifer attached, showing that they opened the email and making their addresses more valuable to the spammer. Squid doesn't match my subdomains

If you are using Squid-2.4 or later then keep in mind that dstdomain acls uses different syntax for exact host matches and entire domain matches. www.example.com matches the exact host www.example.com, while .example.com matches the entire domain example.com (including example.com alone)

There is also subtle issues if your dstdomain ACLs contains matches for both an exact host in a domain and the whole domain where both are in the same domain (i.e. both www.example.com and .example.com). Depending on how your data is ordered this may cause only the most specific of these (e.g. www.example.com) to be used.

NOTE: Current Squid versions (as of Squid-2.4) will warn you when this kind of configuration is used. If your Squid does not warn you while reading the configuration file you do not have the problem described below. Also the configuration here uses the dstdomain syntax of Squid-2.1 or earlier.. (2.2 and later needs to have domains prefixed by a dot)

There is a subtle problem with domain-name based access controls when a single ACL element has an entry that is a subdomain of another entry. For example, consider this list: acl FOO dstdomain boulder.co.us vail.co.us co.us

In the first place, the above list is simply wrong because the first two ( The problem stems from the data structure used to index domain names in an access control list. Squid uses The problem is that it is wrong to say that For example, if you said that similarly, if you said that The bottom line is that you can't have one entry that is a subdomain of another. Squid-2.2 will warn you if it detects this condition. Why does Squid deny some port numbers?

It is dangerous to allow Squid to connect to certain port numbers. For example, it has been demonstrated that someone can use Squid as an SMTP (email) relay. As I'm sure you know, SMTP relays are one of the ways that spammers are able to flood our mailboxes. To prevent mail relaying, Squid denies requests when the URL port number is 25. Other ports should be blocked as well, as a precaution.

There are two ways to filter by port number: either allow specific ports, or deny specific ports. By default, Squid does the first. This is the ACL entry that comes in the default acl Safe_ports port 80 21 443 563 70 210 1025-65535 http_access deny !Safe_ports The above configuration denies requests when the URL port number is not in the list. The list allows connections to the standard ports for HTTP, FTP, Gopher, SSL, WAIS, and all non-priveleged ports.

Another approach is to deny dangerous ports. The dangerous port list should look something like: acl Dangerous_ports 7 9 19 22 23 25 53 109 110 119 http_access deny Dangerous_ports ...and probably many others.

Please consult the /etc/services file on your system for a list of known ports and protocols. Does Squid support the use of a database such as mySQL for storing the ACL list?

Note: The information here is current for version 2.2.

No, it does not. How can I allow a single address to access a specific URL?

This example allows only the acl special_client src acl special_url url_regex ^http://www.squid-cache.org/Doc/FAQ/$ http_access allow special_client special_url http_access deny special_url How can I allow some clients to use the cache at specific times?

Let's say you have two workstations that should only be allowed access to the Internet during working hours (8:30 - 17:30). You can use something like this: acl FOO src acl WORKING time MTWHF 08:30-17:30 http_access allow FOO WORKING http_access deny FOO How can I allow some users to use the cache at specific times?

acl USER1 proxy_auth Dick acl USER2 proxy_auth Jane acl DAY time 06:00-18:00 http_access allow USER1 DAY http_access deny USER1 http_access allow USER2 !DAY http_access deny USER2 Problems with IP ACL's that have complicated netmasks

Note: The information here is current for version 2.3.

The following ACL entry gives inconsistent or unexpected results: acl restricted src The reason is that IP access lists are stored in ``splay'' tree data structures. These trees require the keys to be sortable. When you use a complicated, or non-standard, netmask (, it confuses the function that compares two address/mask pairs.

The best way to fix this problem is to use separate ACL names for each ACL value. For example, change the above to: acl restricted1 src acl restricted2 src

Then, of course, you'll have to rewrite your Can I set up ACL's based on MAC address rather than IP?

Yes, for some operating systes. Squid calls these ``ARP ACLs'' and they are supported on Linux, Solaris, and probably BSD variants.

NOTE: Squid can only determine the MAC address for clients that are on the same subnet. If the client is on a different subnet, then Squid can not find out its MAC address.

To use ARP (MAC) access controls, you first need to compile in the optional code. Do this with the % ./configure --enable-arp-acl ... % make clean % make If src/acl.c doesn't compile, then ARP ACLs are probably not supported on your system.

If everything compiles, then you can add some ARP ACL lines to your acl M1 arp 01:02:03:04:05:06 acl M2 arp 11:12:13:14:15:16 http_access allow M1 http_access allow M2 http_access deny all Debugging ACLs

See and . Can I limit the number of connections from a client?

Yes, use the acl losers src acl 5CONN maxconn 5 http_access deny 5CONN losers

Given the above configuration, when a client whose source IP address is in the subnet tries to establish 6 or more connections at once, Squid returns an error page. Unless you use the The Note, the Also note that you could use I'm trying to deny In Squid-2.3 we changed the way that Squid matches subdomains. There is a difference between acl yuck dstdomain .foo.com http_access deny yuck I want to customize, or make my own error messages.

You can customize the existing error messages as described in . You can also create new error messages and use these in conjunction with the For example, lets say you want your users to see a special message when they request something that matches your pornography list. First, create a file named ERR_NO_PORNO in the /usr/local/squid/etc/errors directory. That file might contain something like this: <p> Our company policy is to deny requests to known porno sites. If you feel you've received this message in error, please contact the support staff (support@this.company.com, 555-1234).

Next, set up your access controls as follows: acl porn url_regex "/usr/local/squid/etc/porno.txt" deny_info ERR_NO_PORNO porn http_access deny porn (additional http_access lines ...) I want to use local time zone in error messages

Squid by defaults uses GMT as timestamp in all geenrated error messages. This to allow the cache to participate in a hierarchy of caches in different timezones without risking confusion about what the time is.

To change the timestamp in Squid generated error messages you must change the Squid signature. See . The signature by defaults uses %T as timestamp, but if you like then you can use %t instead for a timestamp using local time zone. Troubleshooting Why am I getting ``Proxy Access Denied?''

You may need to set up the for information about that.

If httpd_accel_with_proxy on Alternately, you may have misconfigured one of your ACLs. Check the I can't get The I get If the HTTP port number is wrong but the ICP port is correct you will send ICP queries correctly and the ICP replies will fool your cache into thinking the configuration is correct but large objects will fail since you don't have the correct HTTP port for the sibling in your Running out of filedescriptors

If you see the Linux

Henrik has a page.

You also might want to have a look at by

If your kernel version is 2.2.x or greater, you can read and write the maximum number of file handles and/or inodes simply by accessing the special files: /proc/sys/fs/file-max /proc/sys/fs/inode-max So, to increase your file descriptor limit: echo 3072 > /proc/sys/fs/file-max

If your kernel version is between 2.0.35 and 2.1.x (?), you can read and write the maximum number of file handles and/or inodes simply by accessing the special files: /proc/sys/kernel/file-max /proc/sys/kernel/inode-max

While this does increase the current number of file descriptors, Squid's /usr/include/linux/limits.h. Solaris

Add the following to your /etc/system file to increase your maximum file descriptors per process:

set rlim_fd_max = 4096

Next you should re-run the configure script in the top directory so that it finds the new value. If it does not find the new limit, then you might try editing include/autoconf.h and setting include/autoconf.h is created from autoconf.h.in every time you run configure. Thus, if you edit it by hand, you might lose your changes later on.

If you have a very old version of Squid (1.1.X), and you want to use more than 1024 descriptors, then you must edit src/Makefile and enable advises that you should NOT change the default soft limit (FreeBSD

by How do I check my maximum filedescriptors?

Do How do I increase them? sysctl -w kern.maxfiles=XXXX sysctl -w kern.maxfilesperproc=XXXX Warning: You probably want What is the upper limit?

I don't think there is a formal upper limit inside the kernel. All the data structures are dynamically allocated. In practice there might be unintended metaphenomena (kernel spending too much time searching tables, for example). General BSD

For most BSD-derived systems (SunOS, 4.4BSD, OpenBSD, FreeBSD, NetBSD, BSD/OS, 386BSD, Ultrix) you can also use the ``brute force'' method to increase these values in the kernel (requires a kernel rebuild): How do I check my maximum filedescriptors?

Do How do I increase them the easy way?

One way is to increase the value of the Is there a more precise method?

Another way is to find the Here are a few examples which should lead you in the right direction: SunOS

Change the value of by altering this equation: int nfile = 16 * (NPROC + 16 + MAXUSERS) / 10 + 64; Where #define NPROC (10 + 16 * MAXUSERS) FreeBSD (from the 2.1.6 kernel)

Very similar to SunOS, edit /usr/src/sys/conf/param.c and alter the relationship between maxfiles and maxfilesperproc variables: int maxfiles = NPROC*2; int maxfilesperproc = NPROC*2; Where NPROC is defined by: #define NPROC (20 + 16 * MAXUSERS) The per-process limit can also be adjusted directly in the kernel configuration file with the following directive: options OPEN_MAX=128 BSD/OS (from the 2.1 kernel)

Edit /usr/src/sys/conf/param.c and adjust the maxfiles math here: int maxfiles = 3 * (NPROC + MAXUSERS) + 80; Where NPROC is defined by: #define NPROC (20 + 16 * MAXUSERS) You should also set the OPEN_MAX value in your kernel configuration file to change the per-process limit. Reconfigure afterwards

cd squid-1.1.x make realclean ./configure --prefix=/usr/local/squid make What are these strange lines about removing objects?

For example: 97/01/23 22:31:10| Removed 1 of 9 objects from bucket 3913 97/01/23 22:33:10| Removed 1 of 5 objects from bucket 4315 97/01/23 22:35:40| Removed 1 of 14 objects from bucket 6391 These log entries are normal, and do not indicate that Consult your cache information page in Storage LRU Expiration Age: 364.01 days Objects which have not been used for that amount of time are removed as a part of the regular maintenance. You can set an upper limit on the Can I change a Windows NT FTP server to list directories in Unix format?

Why, yes you can! Select the following menus: Start Programs Microsoft Internet Server (Common) Internet Service Manager

This will bring up a box with icons for your various services. One of them should be a little ftp ``folder.'' Double click on this.

You will then have to select the server (there should only be one) Select that and then choose ``Properties'' from the menu and choose the ``directories'' tab along the top.

There will be an option at the bottom saying ``Directory listing style.'' Choose the ``Unix'' type, not the ``MS-DOS'' type.

--Oskar Pearson <oskar@is.co.za> Why am I getting ``Ignoring MISS from non-peer x.x.x.x?''

You are receiving ICP MISSes (via UDP) from a parent or sibling cache whose IP address your cache does not know about. This may happen in two situations.

If the peer is multihomed, it is sending packets out an interface which is not advertised in the DNS. Unfortunately, this is a configuration problem at the peer site. You can tell them to either add the IP address interface to their DNS, or use Squid's "udp_outgoing_address" option to force the replies out a specific interface. For example:

udp_outgoing_address proxy.parent.com cache_peer proxy.parent.com parent 3128 3130 You can also see this warning when sending ICP queries to multicast addresses. For security reasons, Squid requires your configuration to list all other caches listening on the multicast group address. If an unknown cache listens to that address and sends replies, your cache will log the warning message. To fix this situation, either tell the unknown cache to stop listening on the multicast address, or if they are legitimate, add them to your configuration file. DNS lookups for domain names with underscores (_) always fail.

The standards for naming hosts (, ) do not allow underscores in domain names: A "name" (Net, Host, Gateway, or Domain name) is a text string up to 24 characters drawn from the alphabet (A-Z), digits (0-9), minus sign (-), and period (.). The resolver library that ships with recent versions of BIND enforces this restriction, returning an error for any host with underscore in the hostname. The best solution is to complain to the hostmaster of the offending site, and ask them to rename their host.

See also the .

Some people have noticed that implies that underscores Why does Squid say: ``Illegal character in hostname; underscores are not allowed?'

See the above question. The underscore character is not valid for hostnames.

Some DNS resolvers allow the underscore, so yes, the hostname might work fine when you don't use Squid.

To make Squid allow underscores in hostnames, re-run the configure script with this option: % ./configure --enable-underscores ... and then recompile: % make clean % make Why am I getting access denied from a sibling cache?

The answer to this is somewhat complicated, so please hold on. .

An ICP query does not include any parent or sibling designation, so the receiver really has no indication of how the peer cache is configured to use it. This issue becomes important when a cache is willing to serve cache hits to anyone, but only handle cache misses for its paying users or customers. In other words, whether or not to allow the request depends on if the result is a hit or a miss. To accomplish this, Squid acquired the The necessity of ``miss access'' makes life a little bit complicated, and not only because it was awkward to implement. Miss access means that the ICP query reply must be an extremely accurate prediction of the result of a subsequent HTTP request. Ascertaining this result is actually very hard, if not impossible to do, since the ICP request cannot convey the full HTTP request. Additionally, there are more types of HTTP request results than there are for ICP. The ICP query reply will either be a hit or miss. However, the HTTP request might result in a `` One serious problem for cache hierarchies is mismatched freshness parameters. Consider a cache In an HTTP/1.0 world, HTTP/1.1 provides numerous request headers to specify freshness requirements, which actually introduces a different problem for cache hierarchies: ICP still does not include any age information, neither in query nor reply. So In the end, the fundamental problem is that the ICP query does not provide enough information to accurately predict whether the HTTP request will be a hit or miss. In fact, the current ICP Internet Draft is very vague on this subject. What does ICP HIT really mean? Does it mean ``I know a little about that URL and have some copy of the object?'' Or does it mean ``I have a valid copy of that object and you are allowed to get it from me?''

So, what can be done about this problem? We really need to change ICP so that freshness parameters are included. Until that happens, the members of a cache hierarchy have only two options to totally eliminate the ``access denied'' messages from sibling caches: Make sure all members have the same Do not use If neither of these is realistic, then the sibling relationship should not exist. Cannot bind socket FD NN to *:8080 (125) Address already in use

This means that another processes is already listening on port 8080 (or whatever you're using). It could mean that you have a Squid process already running, or it could be from another program. To verify, use the netstat -naf inet | grep LISTEN That will show all sockets in the LISTEN state. You might also try netstat -naf inet | grep 8080 If you find that some process has bound to your port, but you're not sure which process it is, you might be able to use the excellent program. It will show you which processes own every open file descriptor on your system. icpDetectClientClose: ERROR xxx.xxx.xxx.xxx: (32) Broken pipe

This means that the client socket was closed by the client before Squid was finished sending data to it. Squid detects this by trying to icpDetectClientClose: FD 135, 255 unexpected bytes

These are caused by misbehaving Web clients attempting to use persistent connections. Squid-1.1 does not support persistent connections. Does Squid work with NTLM Authentication?

will support Microsoft NTLM authentication. However, there are some limits on our support: We cannot proxy connections to a origin server that use NTLM authentication, but we can act as a web accelerator or proxy server and authenticate the client connection using NTLM.

We support NT4, Samba, and Windows 2000 Domain Controllers. For more information see .

Why we cannot proxy NTLM even though we can use it. Quoting from summary at the end of the browser authentication section in : In summary, Basic authentication does not require an implicit end-to-end state, and can therefore be used through a proxy server. Windows NT Challenge/Response authentication requires implicit end-to-end state and will not work through a proxy server.

Squid transparently passes the NTLM request and response headers between clients and servers. NTLM relies on a single end-end connection (possibly with men-in-the-middle, but a single connection every step of the way. This implies that for NTLM authentication to work at all with proxy caches, the proxy would need to tightly link the client-proxy and proxy-server links, as well as understand the state of the link at any one time. NTLM through a CONNECT might work, but we as far as we know that hasn't been implemented by anyone, and it would prevent the pages being cached - removing the value of the proxy.

NTLM authentication is carried entirely inside the HTTP protocol, but is not a true HTTP authentication protocol and is different from Basic and Digest authentication in many ways. It is dependent on a stateful end-to-end connection which collides with RFC 2616 for proxy-servers to disjoin the client-proxy and proxy-server connections. It is only taking place once per connection, not per request. Once the connection is authenticated then all future requests on the same connection inherities the authentication. The connection must be reestablished to set up other authentication or re-identify the user. This too collides with RFC 2616 where authentication is defined as a property of the HTTP messages, not connections.

The reasons why it is not implemented in Netscape is probably: It is very specific for the Windows platform It is not defined in any RFC or even internet draft. The protocol has several shortcomings, where the most apparent one is that it cannot be proxied. There exists an open internet standard which does mostly the same but without the shortcomings or platform dependencies: . The This message was received at If you have only one parent, configured as: cache_peer xxxx parent 3128 3130 no-query default nothing is sent to the parent; neither UDP packets, nor TCP connections.

Simply adding acl all src never_direct allow all ``Hot Mail'' complains about: Intrusion Logged. Access denied.

``Hot Mail'' is proxy-unfriendly and requires all requests to come from the same IP address. You can fix this by adding to your hierarchy_stoplist hotmail.com My Squid becomes very slow after it has been running for some time.

This is most likely because Squid is using more memory than it should be for your system. When the Squid process becomes large, it experiences a lot of paging. This will very rapidly degrade the performance of Squid. Memory usage is a complicated problem. There are a number of things to consider.

Then, examine the Cache Manager Number of HTTP requests received: 121104 Page faults with physical i/o: 16720 Note, if your system does not have the Divide the number of page faults by the number of connections. In this case 16720/121104 = 0.14. Ideally this ratio should be in the 0.0 - 0.1 range. It may be acceptable to be in the 0.1 - 0.2 range. Above that, however, and you will most likely find that Squid's performance is unacceptably slow.

If the ratio is too high, you will need to make some changes to .

See also . WARNING: Failed to start 'dnsserver'

This could be a permission problem. Does the Squid userid have permission to execute the You might also try testing > echo oceana.nlanr.net | ./dnsserver Should produce something like: $name oceana.nlanr.net $h_name oceana.nlanr.net $h_len 4 $ipcount 1 $aliascount 0 $ttl 82067 $end Sending in Squid bug reports

Bug reports for Squid should be registered in our . Any bug report must include The Squid version Your Operating System type and version A clear description of the bug symptoms. If your Squid crashes the report must include a as described below

Please note that bug reports are only processed if they can be reproduced or identified in the current STABLE or development versions of Squid. If you are running an older version of Squid the first response will be to ask you to upgrade unless the developer who looks at your bug report immediately can identify that the bug also exists in the current versions. It should also be noted that any patches provided by the Squid developer team will be to the current STABLE version even if you run an older version. crashes and core dumps

There are two conditions under which squid will exit abnormally and generate a coredump. First, a SIGSEGV or SIGBUS signal will cause Squid to exit and dump core. Second, many functions include consistency checks. If one of those checks fail, Squid calls abort() to generate a core dump.

Many people report that Squid doesn't leave a coredump anywhere. This may be due to one of the following reasons: Resource Limits. The shell has limits on the size of a coredump file. You may need to increase the limit. sysctl options. On FreeBSD, you won't get a coredump from programs that call setuid() and/or setgid() (like Squid sometimes does) unless you enable this option: # sysctl -w kern.sugid_coredump=1 No debugging symbols. The Squid binary must have debugging symbols in order to get a meaningful coredump. Threads and Linux. On Linux, threaded applications do not generate core dumps. When you use the aufs cache_dir type, it uses threads and you can't get a coredump. It did leave a coredump file, you just can't find it.

/etc/login.conf file on FreeBSD and maybe other BSD systems.

To change the coredumpsize limit you might use a command like: limit coredumpsize unlimited or limits coredump unlimited

% nm /usr/local/squid/bin/squid | head The binary has debugging symbols if you see gobbledegook like this: 0812abec B AS_tree_head 080a7540 D AclMatchedName 080a73fc D ActionTable 080908a4 r B_BYTES_STR 080908bc r B_GBYTES_STR 080908ac r B_KBYTES_STR 080908b4 r B_MBYTES_STR 080a7550 D Biggest_FD 08097c0c R CacheDigestHashFuncCount 08098f00 r CcAttrs There are no debugging symbols if you see this instead: /usr/local/squid/bin/squid: no symbols Debugging symbols may have been removed by your The The first The current directory when Squid was started Recent versions of Squid report their current directory after starting, so look there first: 2000/03/14 00:12:36| Set Current Directory to /usr/local/squid/cache If you cannot find a core file, then either Squid does not have permission to write in its current directory, or perhaps your shell limits are preventing the core file from being written.

Often you can get a coredump if you run Squid from the command line like this (csh shells and clones): % limit core un % /usr/local/squid/bin/squid -NCd1

Once you have located the core dump file, use a debugger such as tirana-wessels squid/src 270% gdb squid /T2/Cache/core GDB is free software and you are welcome to distribute copies of it under certain conditions; type "show copying" to see the conditions. There is absolutely no warranty for GDB; type "show warranty" for details. GDB 4.15.1 (hppa1.0-hp-hpux10.10), Copyright 1995 Free Software Foundation, Inc... Core was generated by `squid'. Program terminated with signal 6, Aborted. [...] (gdb) where #0 0xc01277a8 in _kill () #1 0xc00b2944 in _raise () #2 0xc007bb08 in abort () #3 0x53f5c in __eprintf (string=0x7b037048 "", expression=0x5f

, line=8, filename=0x6b
) #4 0x29828 in fd_open (fd=10918, type=3221514150, desc=0x95e4 "HTTP Request") at fd.c:71 #5 0x24f40 in comm_accept (fd=2063838200, peer=0x7b0390b0, me=0x6b) at comm.c:574 #6 0x23874 in httpAccept (sock=33, notused=0xc00467a6) at client_side.c:1691 #7 0x25510 in comm_select_incoming () at comm.c:784 #8 0x25954 in comm_select (sec=29) at comm.c:1052 #9 0x3b04c in main (argc=1073745368, argv=0x40000dd8) at main.c:671

If possible, you might keep the coredump file around for a day or two. It is often helpful if we can ask you to send additional debugger output, such as the contents of some variables. But please note that a core file is only useful if paired with the exact same binary as generated the corefile. If you recompile Squid then any coredumps from previous versions will be useless unless you have saved the corresponding Squid binaries, and any attempts to analyze such coredumps will most certainly give misleading information about the cause to the crash.

If you CANNOT get Squid to leave a core file for you then one of the following approaches can be used

First alternative is to start Squid under the contol of GDB % gdb /path/to/squid handle SIGPIPE pass nostop noprint run -DNYCd3 [wait for crash] backtrace quit

The drawback from the above is that it isn't really suitable to run on a production system as Squid then won't restart automatically if it crashes. The good news is that it is fully possible to automate the process above to automatically get the stack trace and then restart Squid. Here is a short automated script that should work: #!/bin/sh trap "rm -f $$.gdb" 0 cat <$$.gdb handle SIGPIPE pass nostop noprint run -DNYCd3 backtrace quit EOF while sleep 2; do gdb -x $$.gdb /path/to/squid 2>&1 | tee -a squid.out done

Other options if the above cannot be done is to:

a) Build Squid with the --enable-stacktraces option, if support exists for your OS (exists for Linux glibc on Intel, and Solaris with some extra libraries which seems rather impossible to find these days..)

b) Run Squid using the "catchsegv" tool. (Linux glibc Intel)

but these approaches does not by far provide as much details as using gdb. Debugging Squid

If you believe you have found a non-fatal bug (such as incorrect HTTP processing) please send us a section of your cache.log with debugging to demonstrate the problem. The cache.log file can become very large, so alternatively, you may want to copy it to an FTP or HTTP server where we can download it.

It is very simple to enable full debugging on a running squid process. Simply use the % ./squid -k debug This causes every To enable selective debugging (e.g. for one source file only), you need to edit doc/debug-levels.txt (correctly renamed to debug_options ALL,1 28,9 Then you have to restart or reconfigure Squid.

Once you have the debugging captured to FATAL: ipcache_init: DNS name lookup tests failed

Squid normally tests your system's DNS configuration before it starts server requests. Squid tries to resolve some common DNS names, as defined in the your DNS nameserver is unreachable or not running. your /etc/resolv.conf file may contain incorrect information. your /etc/resolv.conf file may have incorrect permissions, and may be unreadable by Squid.

To disable this feature, use the Note, Squid does NOT use the FATAL: Failed to make swap directory /var/spool/cache: (13) Permission denied

Starting with version 1.1.15, we have required that you first run squid -z to create the swap directories on your filesystem. If you have set the # mkdir /var/spool/cache # chown /var/spool/cache # squid -z

Alternatively, if the directory already exists, then your operating system may be returning ``Permission Denied'' instead of ``File Exists'' on the mkdir() system call. This by should fix it. FATAL: Cannot open HTTP Port

Either (1) the Squid userid does not have permission to bind to the port, or (2) some other process has bound itself to the port. Remember that root privileges are required to open port numbers less than 1024. If you see this message when using a high port number, or even when starting Squid as root, then the port has already been opened by another process. Maybe you are running in the HTTP Accelerator mode and there is already a HTTP server running on port 80? If you're really stuck, install the way cool utility to show you which process has your port in use. FATAL: All redirectors have exited!

This is explained in the . FATAL: file_map_allocate: Exceeded filemap limit

See the next question. FATAL: You've run out of swap file numbers.

Note: The information here applies to version 2.2 and earlier.

Squid keeps an in-memory bitmap of disk files that are available for use, or are being used. The size of this bitmap is determined at run name, based on two things: the size of your cache, and the average (mean) cache object size. The size of your cache is specified in squid.conf, on the When allocating the bitmaps, Squid allocates this many bits: 2 * cache_size / store_avg_object_size So, if you exactly specify the correct average object size, Squid should have 50% filemap bits free when the cache is full. You can see how many filemap bits are being used by looking at the 'storedir' cache manager page. It looks like this: Store Directory #0: /usr/local/squid/cache First level subdirectories: 4 Second level subdirectories: 4 Maximum Size: 1024000 KB Current Size: 924837 KB Percent Used: 90.32% Filemap bits in use: 77308 of 157538 (49%) Flags:

Now, if you see the ``You've run out of swap file numbers'' message, then it means one of two things: You've found a Squid bug. Your cache's average file size is much smaller than the 'store_avg_object_size' value. To check the average file size of object currently in your cache, look at the cache manager 'info' page, and you will find a line like: Mean Object Size: 11.96 KB

To make the warning message go away, set 'store_avg_object_size' to that value (or lower) and then restart Squid. I am using up over 95% of the filemap bits?!!

Note: The information here is current for version 2.3

Calm down, this is now normal. Squid now dynamically allocates filemap bits based on the number of objects in your cache. You won't run out of them, we promise. FATAL: Cannot open /usr/local/squid/logs/access.log: (13) Permission denied

In Unix, things like To find out who owns a file, use the % ls -l /usr/local/squid/logs/access.log

A process is normally owned by the user who starts it. However, Unix sometimes allows a process to change its owner. If you specified a value for the If all this is confusing, then you probably should not be running Squid until you learn some more about Unix. As a reference, I suggest . When using a username and password, I can not access some files.

If I try by way of a test, to access ftp://username:password@ftpserver/somewhere/foo.tar.gz I get somewhere/foo.tar.gz: Not a directory.

Use this URL instead: ftp://username:password@ftpserver/%2fsomewhere/foo.tar.gz pingerOpen: icmp_sock: (13) Permission denied

This means your % su # make install-pinger or # chown root /usr/local/squid/bin/pinger # chmod 4755 /usr/local/squid/bin/pinger What is a forwarding loop?

A forwarding loop is when a request passes through one proxy more than once. You can get a forwarding loop if a cache forwards requests to itself. This might happen with interception caching (or server acceleration) configurations. a pair or group of caches forward requests to each other. This can happen when Squid uses ICP, Cache Digests, or the ICMP RTT database to select a next-hop cache.

Forwarding loops are detected by examining the NOTE: Squid may report a forwarding loop if a request goes through two caches that have the same When Squid detects a forwarding loop, it is logged to the One way to reduce forwarding loops is to change a Another way is to use # Our parent caches cache_peer A.example.com parent 3128 3130 cache_peer B.example.com parent 3128 3130 cache_peer C.example.com parent 3128 3130 # An ACL list acl PEERS src A.example.com acl PEERS src B.example.com acl PEERS src C.example.com # Prevent forwarding loops cache_peer_access A.example.com allow !PEERS cache_peer_access B.example.com allow !PEERS cache_peer_access C.example.com allow !PEERS The above configuration instructs squid to NOT forward a request to parents A, B, or C when a request is received from any one of those caches. accept failure: (71) Protocol error

This error message is seen mostly on Solaris systems. gives a great explanation: Error 71 [EPROTO] is an obscure way of reporting that clients made it onto your server's TCP incoming connection queue but the client tore down the connection before the server could accept it. I.e. your server ignored its clients for too long. We've seen this happen when we ran out of file descriptors. I guess it could also happen if something made squid block for a long time. storeSwapInFileOpened: ... Size mismatch

Got these messages in my cache log - I guess it means that the index contents do not match the contents on disk. 1998/09/23 09:31:30| storeSwapInFileOpened: /var/cache/00/00/00000015: Size mismatch: 776(fstat) != 3785(object) 1998/09/23 09:31:31| storeSwapInFileOpened: /var/cache/00/00/00000017: Size mismatch: 2571(fstat) != 4159(object)

What does Squid do in this case?

NOTE, these messages are specific to Squid-2. These happen when Squid reads an object from disk for a cache hit. After it opens the file, Squid checks to see if the size is what it expects it should be. If the size doesn't match, the error is printed. In this case, Squid does not send the wrong object to the client. It will re-fetch the object from the source. Why do I get fwdDispatch: Cannot retrieve 'https://www.buy.com/corp/ordertracking.asp'

These messages are caused by buggy clients, mostly Netscape Navigator. What happens is, Netscape sends an HTTPS/SSL request over a persistent HTTP connection. Normally, when Squid gets an SSL request, it looks like this: CONNECT www.buy.com:443 HTTP/1.0 Then Squid opens a TCP connection to the destination host and port, and the With this client bug, however, Squid receives a request like this: GET https://www.buy.com/corp/ordertracking.asp HTTP/1.0 Accept: */* User-agent: Netscape ... ... Now, all of the headers, and the message body have been sent, Note, this browser bug does represent a security risk because the browser is sending sensitive information unencrypted over the network. Squid can't access URLs like http://3626046468/ab2/cybercards/moreinfo.html

by Dave J Woolley (DJW at bts dot co dot uk)

These are illegal URLs, generally only used by illegal sites; typically the web site that supports a spammer and is expected to survive a few hours longer than the spamming account.

Their intention is to: confuse content filtering rules on proxies, and possibly some browsers' idea of whether they are trusted sites on the local intranet; confuse whois (?); make people think they are not IP addresses and unknown domain names, in an attempt to stop them trying to locate and complain to the ISP.

Any browser or proxy that works with them should be considered a security risk.

has this to say about the hostname part of a URL: The fully qualified domain name of a network host, or its IP address as a set of four decimal digit groups separated by ".". Fully qualified domain names take the form as described in Section 3.5 of RFC 1034 [13] and Section 2.1 of RFC 1123 [5]: a sequence of domain labels separated by ".", each domain label starting and ending with an alphanumerical character and possibly also containing "-" characters. The rightmost domain label will never start with a digit, though, which syntactically distinguishes all domain names from the IP addresses. I get a lot of ``URI has whitespace'' error messages in my cache log, what should I do?

Whitespace characters (space, tab, newline, carriage return) are not allowed in URI's and URL's. Unfortunately, a number of Web services generate URL's with whitespace. Of course your favorite browser silently accomodates these bad URL's. The servers (or people) that generate these URL's are in violation of Internet standards. The whitespace characters should be encoded.

If you want Squid to accept URL's with whitespace, you have to decide how to handle them. There are four choices that you can set with the DENY: The request is denied with an ``Invalid Request'' message. This is the default. ALLOW: The request is allowed and the URL remains unchanged. ENCODE: The whitespace characters are encoded according to . This can be considered a violation of the HTTP specification. CHOP: The URL is chopped at the first whitespace character and then processed normally. This also can be considered a violation of HTTP. commBind: Cannot bind socket FD 5 to (49) Can't assign requested address

This likely means that your system does not have a loopback network device, or that device is not properly configured. All Unix systems should have a network device named % ifconfig lo0 The result should look something like: lo0: flags=8049 mtu 16384 inet netmask 0xff000000

If you use FreeBSD, see . Unknown cache_dir type '/var/squid/cache'

The format of the cache_dir ufs /var/squid/cache ... unrecognized: 'cache_dns_program /usr/local/squid/bin/dnsserver'

As of Squid 2.3, the default is to use internal DNS lookup code. The If you want to use external DNS lookups, with the --disable-internal-dns Is Sort of. As of Squid 2.3, the default is to use internal DNS lookup code. The See if the Configure squid with --disable-internal-dns to use the external dnsservers. Enhance src/dns_internal.c to understand the /etc/resolv.conf. What does sslReadClient: FD 14: read failure: (104) Connection reset by peer mean?

``Connection reset by peer'' is an error code that Unix operating systems sometimes return for Connection reset means that the other host, the peer, sent us a RESET packet on a TCP connection. A host sends a RESET when it receives an unexpected packet for a nonexistent connection. For example, if one side sends data at the same time that the other side closes a connection, when the other side receives the data it may send a reset back.

The fact that these messages appear in Squid's log might indicate a problem, such as a broken origin server or parent cache. On the other hand, they might be ``normal,'' especially since some applications are known to force connection resets rather than a proper close.

You probably don't need to worry about them, unless you receive a lot of user complaints relating to SSL sites.

notes that if the server is running a Microsoft TCP stack, clients receive RST segments whenever the listen queue overflows. In other words, if the server is really busy, new connections receive the reset message. This is contrary to rational behaviour, but is unlikely to change. What does Connection refused mean?

This is an error message, generated by your operating system, in response to a Its quite easy to generate this error on your own. Simply telnet to a random, high numbered port: % telnet localhost 12345 Trying telnet: Unable to connect to remote host: Connection refused It happens because there is no server listening for connections on port 12345.

When you see this in response to a URL request, it probably means the origin server web site is temporarily down. It may also mean that your parent cache is down, if you have one. squid: ERROR: no running copy

You may get this message when you run commands like This error message usually means that the If you accidentally removed the PID file, there are two ways to get it back. run bender-wessels % ps ax | grep squid 83617 ?? Ss 0:00.00 squid -s 83619 ?? S 0:00.48 (squid) -s (squid) You want the second process id, 83619 in this case. Create the PID file and put the process id number there. For example: echo 83619 > /usr/local/squid/logs/squid.pid Use the above technique to find the Squid process id. Send the process a HUP signal, which is the same as kill -HUP 83619 The reconfigure process creates a new PID file automatically. FATAL: getgrnam failed to find groupid for effective group 'nogroup'

You are probably starting Squid as root. Squid is trying to find a group-id that doesn't have any special priveleges that it will run as. The default is /etc/group. There is a good chance that ``Unsupported Request Method and Protocol'' for Note: The information here is current for version 2.3.

This is correct. Squid does not know what to do with an Normally, when you type an The browser opens an SSL connection directly to the origin server. The browser tunnels the request through Squid with the

The and (expired). Squid uses 100% CPU

There may be many causes for this.

Andrew Doroshenko reports that removing /dev/null, or mounting a filesystem with the nodev option, can cause Squid to use 100% of CPU. His suggested solution is to ``touch /dev/null.'' Webmin's Mikael Andersson reports that clicking on Webmin's Joe Cooper reports this to be caused by SSL problems in some browsers (mainly Netscape 6.x/Mozilla) if your Webmin is SSL enabled. Try with another browser such as Netscape 4.x or Microsoft IE, or disable SSL encryption in Webmin. Segment Violation at startup or upon first request

Some versions of GCC (notably 2.95.1 through 2.95.4 at least) have bugs with compiler optimization. These GCC bugs may cause NULL pointer accesses in Squid, resulting in a ``FATAL: Received Segment Violation...dying'' message and a core dump.

You can work around these GCC bugs by disabling compiler optimization. The best way to do that is start with a clean source tree and set the CC options specifically: % cd squid-x.y % make distclean % setenv CFLAGS='-g -Wall' % ./configure ...

To check that you did it right, you can search for AC_CFLAGS in src/Makefile: % grep AC_CFLAGS src/Makefile AC_CFLAGS = -g -Wall Now when you recompile, GCC won't try to optimize anything: % make Making all in lib... gcc -g -Wall -I../include -I../include -c rfc1123.c ...etc...

NOTE: some people worry that disabling compiler optimization will negatively impact Squid's performance. The impact should be negligible, unless your cache is really busy and already runs at a high CPU usage. For most people, the compiler optimization makes little or no difference at all. urlParse: Illegal character in hostname 'proxy.mydomain.com:8080proxy.mydomain.com'

By Yomler of fnac.net

A combination of a bad configuration of Internet Explorer and any application which use the cydoor DLLs will produce the entry in the log. See for a complete list.

The bad configuration of IE is the use of a active configuration script (proxy.pac) and an active or inactive, but filled proxy settings. IE will only use the proxy.pac. Cydoor aps will use both and will generate the errors.

Disabling the old proxy settings in IE is not enought, you should delete them completely and only use the proxy.pac for example. Requests for international domain names does not work

By Henrik Nordström

Some people have asked why requests for domain names using national symbols as "supported" by the certain domain registrars does not work in Squid. This is because there as of yet is no standard on how to manage national characters in the current Internet protocols such as HTTP or DNS. The current Internet standards is very strict on what is an acceptable hostname and only accepts A-Z a-z 0-9 and - in Internet hostname labels. Anything outside this is outside the current Internet standards and will cause interoperability issues such as the problems seen with such names and Squid.

When there is a consensus in the DNS and HTTP standardization groups on how to handle international domain names Squid will be changed to support this if any changes to Squid will be required.

If you are interested in the progress of the standardization process for international domain names please see the IETF IDN working group's . Why do I sometimes get ``Zero Sized Reply''?

This happens when Squid makes a TCP connection to an origin server, but for some reason, the connection is closed before Squid reads any data. Depending on various factors, Squid may be able to retry the request again. If you see the ``Zero Sized Reply'' error message, it means that Squid was unable to retry, or that all retry attempts also failed.

What causes a connection to close prematurely? It could be a number of things, including: An overloaded origin server. TCP implementation/interoperability bugs. Race conditions with HTTP persistent connections. Buggy or misconfigured NAT boxes, firewalls, and load-balancers. Denial of service attacks.

You may be able to use Some users believe the problem is caused by very large cookies. One user reports that his Zero Sized Reply problem went away when he told Internet Explorer to not accept third-party cookies.

Here are some things you can try to reduce the occurance of the Zero Sized Reply error: Delete or rename your cookie file and configure your browser to prompt you before accepting any new cookies. Disable HTTP persistent connections with the Disable any advanced TCP features on the Squid system. Disable ECN on Linux with echo 0 > /proc/sys/net/ipv4/tcp_ecn/. Upgrade to Squid-2.5.STABLE4 or later to work around a Host header related bug in Cisco PIX HTTP inspection. The Cisco PIX firewall wrongly assumes the Host header can be found in the first packet of the request.

If this error causes serious problems for you and the above does not help, Squid developers would be happy to help you uncover the problem. However, we will require high-quality debugging information from you, such as If you want to make Squid give the Zero Sized error on demand, you can use the short C program below. Simply compile and start the program on a system that doesn't already have a server running on port 80. Then try to connect to this fake server through Squid: #include #include #include #include #include #include #include #include int main(int a, char **b) { struct sockaddr_in S; int s,t,x; s = socket(PF_INET, SOCK_STREAM, 0); assert(s > 0); memset(&S, '\0', sizeof(S)); S.sin_family = AF_INET; S.sin_port = htons(80); x = bind(s, (struct sockaddr *) &S, sizeof(S)); assert(x == 0); x = listen(s, 10); assert(x == 0); while (1) { struct sockaddr_in F; int fl = sizeof(F); t = accept(s, (struct sockaddr *) &F, &fl); fprintf(stderr, "accpeted FD %d from %s:%d\n", t, inet_ntoa(F.sin_addr), (int)ntohs(F.sin_port)); close(t); fprintf(stderr, "closed FD %d\n", t); } return 0; } How does Squid work?

An Internet Object is a file, document or response to a query for an Internet service such as FTP, HTTP, or gopher. A client requests an Internet object from a caching proxy; if the object is not already cached, the proxy server fetches the object (either from the host specified in the URL or from a parent or sibling cache) and delivers it to the client. What is the ICP protocol?

ICP is a protocol used for communication among squid caches. The ICP protocol is defined in two Internet RFC's. describes the protocol itself, while describes the application of ICP to hierarchical Web caching.

ICP is primarily used within a cache hierarchy to locate specific objects in sibling caches. If a squid cache does not have a requested document, it sends an ICP query to its siblings, and the siblings respond with ICP replies indicating a ``HIT'' or a ``MISS.'' The cache then uses the replies to choose from which cache to resolve its own MISS.

ICP also supports multiplexed transmission of multiple object streams over a single TCP connection. ICP is currently implemented on top of UDP. Current versions of Squid also support ICP via multicast. What is the The gethostbyname(3) function blocks the calling process until the DNS query is completed.

Squid must use non-blocking I/O at all times, so DNS lookups are implemented external to the main process. The What is the The FTP PUT's don't work!

FTP PUT should work with Squid-2.0 and later versions. If you are using Squid-1.1, then you need to upgrade before PUT will work. What is a cache hierarchy? What are parents and siblings?

A cache hierarchy is a collection of caching proxy servers organized in a logical parent/child and sibling arrangement so that caches closest to Internet gateways (closest to the backbone transit entry-points) act as parents to caches at locations farther from the backbone. The parent caches resolve ``misses'' for their children. In other words, when a cache requests an object from its parent, and the parent does not have the object in its cache, the parent fetches the object, caches it, and delivers it to the child. This ensures that the hierarchy achieves the maximum reduction in bandwidth utilization on the backbone transit links, helps reduce load on Internet information servers outside the network served by the hierarchy, and builds a rich cache on the parents so that the other child caches in the hierarchy will obtain better ``hit'' rates against their parents.

In addition to the parent-child relationships, squid supports the notion of siblings: caches at the same level in the hierarchy, provided to distribute cache server load. Each cache in the hierarchy independently decides whether to fetch the reference from the object's home site or from parent or sibling caches, using a a simple resolution protocol. Siblings will not fetch an object for another sibling to resolve a cache ``miss.'' What is the Squid cache resolution algorithm?

Send ICP queries to all appropriate siblings Wait for all replies to arrive with a configurable timeout (the default is two seconds). Begin fetching the object upon receipt of the first HIT reply, or Fetch the object from the first parent which replied with MISS (subject to weighting values), or Fetch the object from the source

The algorithm is somewhat more complicated when firewalls are involved.

The What features are Squid developers currently working on?

There are several open issues for the caching project namely more automatic load balancing and (both configured and dynamic) selection of parents, routing, multicast cache-to-cache communication, and better recognition of URLs that are not worth caching.

For our other to-do list items, please see our ``TODO'' file in the recent source distributions.

Prospective developers should review the resources available at the Tell me more about Internet traffic workloads

Workload can be characterized as the burden a client or group of clients imposes on a system. Understanding the nature of workloads is important to the managing system capacity. If you are interested in Internet traffic workloads then NLANR's is a good place to start. What are the tradeoffs of caching with the NLANR cache system?

The NLANR root caches are at the NSF supercomputer centers (SCCs), which are interconnected via NSF's high speed backbone service (vBNS). So inter-cache communication between the NLANR root caches does not cross the Internet.

The benefits of hierarchical caching (namely, reduced network bandwidth consumption, reduced access latency, and improved resiliency) come at a price. Caches higher in the hierarchy must field the misses of their descendents. If the equilibrium hit rate of a leaf cache is 50%, half of all leaf references have to be resolved through a second level cache rather than directly from the object's source. If this second level cache has most of the documents, it is usually still a win, but if higher level caches often don't have the document, or become overloaded, then they could actually increase access latency, rather than reduce it.

Where can I find out more about firewalls?

Please see the information site. What is the ``Storage LRU Expiration Age?''

For example: Storage LRU Expiration Age: 4.31 days

The LRU expiration age is a dynamically-calculated value. Any objects which have not been accessed for this amount of time will be removed from the cache to make room for new, incoming objects. Another way of looking at this is that it would take your cache approximately this many days to go from empty to full at your current traffic levels.

As your cache becomes more busy, the LRU age becomes lower so that more objects will be removed to make room for the new ones. Ideally, your cache will have an LRU age value in the range of at least 3 days. If the LRU age is lower than 3 days, then your cache is probably not big enough to handle the volume of requests it receives. By adding more disk space you could increase your cache hit ratio.

The configuration parameter What is ``Failure Ratio at 1.01; Going into hit-only-mode for 5 minutes''?

Consider a pair of caches named A and B. It may be the case that A can reach B, and vice-versa, but B has poor reachability to the rest of the Internet. In this case, we would like B to recognize that it has poor reachability and somehow convey this fact to its neighbor caches.

Squid will track the ratio of failed-to-successful requests over short time periods. A failed request is one which is logged as ERR_DNS_FAIL, ERR_CONNECT_FAIL, or ERR_READ_ERROR. When the failed-to-successful ratio exceeds 1.0, then Squid will return ICP_MISS_NOFETCH instead of ICP_MISS to neighbors. Note, Squid will still return ICP_HIT for cache hits. Does squid periodically re-read its configuration file?

No, you must send a HUP signal to have Squid re-read its configuration file, including access control lists. An easy way to do this is with the squid -k reconfigure How does An object with swap file number We want to unlink file We have a new object to swap out. It is allocated to the first available file number, which happens to be The

So, the problem is, how can we guarantee that In terms of implementation, the only way to send unlink requests to the Unfortunately there are times when Squid can not use the What is an icon URL?

One of the most unpleasant things Squid must do is generate HTML pages of Gopher and FTP directory listings. For some strange reason, people like to have little In Squid 1.0 and 1.1, we used internal browser icons with names like For Squid 2 we include a set of icons in the source distribution. These icon files are loaded by Squid as cached objects at runtime. Thus, every Squid cache now has its own icons to use in Gopher and FTP listings. Just like other objects available on the web, we refer to the icons with , or Can I make my regular FTP clients use a Squid cache?

Nope, its not possible. Squid only accepts HTTP requests. It speaks FTP on the The very cool will download FTP URLs via Squid (and probably any other proxy cache). Why is the select loop average time so high?

Is there any way to speed up the time spent dealing with select? Cachemgr shows: Select loop called: 885025 times, 714.176 ms avg

This number is NOT how much time it takes to handle filedescriptor I/O. We simply count the number of times select was called, and divide the total process running time by the number of select calls.

This means, on average it takes your cache .714 seconds to check all the open file descriptors once. But this also includes time select() spends in a wait state when there is no I/O on any file descriptors. My relatively idle workstation cache has similar numbers: Select loop called: 336782 times, 715.938 ms avg But my busy caches have much lower times: Select loop called: 16940436 times, 10.427 ms avg Select loop called: 80524058 times, 10.030 ms avg Select loop called: 10590369 times, 8.675 ms avg Select loop called: 84319441 times, 9.578 ms avg How does Squid deal with Cookies?

The presence of Cookies headers in The proper way to deal with is to cache the whole object, With Squid-1.1, we can not filter out specific HTTP headers, so Squid-1.1 does not cache any response which contains a With Squid-2, however, we can filter out specific HTTP headers. But instead of filtering them on the receiving-side, we filter them on the sending-side. Thus, Squid-2 does cache replies with How does Squid decide when to refresh a cached object?

When checking the object freshness, we calculate these values: OBJ_AGE = NOW - OBJ_DATE LM_AGE = OBJ_DATE - OBJ_LASTMOD LM_FACTOR = OBJ_AGE / LM_AGE

These values are compared with the parameters of the URL regular expression

The URL regular expressions are checked in the order listed until a match is found. Then the algorithms below are applied for determining if an object is fresh or stale. Squid-1.1 and Squid-1.NOVM algorithm

if (CLIENT_MAX_AGE) if (OBJ_AGE > CLIENT_MAX_AGE) return STALE if (OBJ_AGE <= CONF_MIN) return FRESH if (EXPIRES) { if (EXPIRES <= NOW) return STALE else return FRESH } if (OBJ_AGE > CONF_MAX) return STALE if (LM_FACTOR < CONF_PERCENT) return FRESH return STALE

has made an excellent showing this process. Squid-2 algorithm

For Squid-2 the refresh algorithm has been slightly modified to give the if (EXPIRES) { if (EXPIRES <= NOW) return STALE else return FRESH } if (CLIENT_MAX_AGE) if (OBJ_AGE > CLIENT_MAX_AGE) return STALE if (OBJ_AGE > CONF_MAX) return STALE if (OBJ_DATE > OBJ_LASTMOD) { if (LM_FACTOR < CONF_PERCENT) return FRESH else return STALE } if (OBJ_AGE <= CONF_MIN) return FRESH return STALE What exactly is a The cachemanager I/O page lists Sometimes reading on the server-side gets ahead of writing to the client-side. Especially if your cache is on a fast network and your clients are connected at modem speeds. Squid-1.1 will read up to 256k (per request) ahead before it starts to defer the server-side reads. Why is my cache's inbound traffic equal to the outbound traffic?

I've been monitoring the traffic on my cache's ethernet adapter an found a behavior I can't explain: the inbound traffic is equal to the outbound traffic. The differences are negligible. The hit ratio reports 40%. Shouldn't the outbound be at least 40% greater than the inbound?


I can't account for the exact behavior you're seeing, but I can offer this advice; whenever you start measuring raw Ethernet or IP traffic on interfaces, you can forget about getting all the numbers to exactly match what Squid reports as the amount of traffic it has sent/received.


Squid is an application - it counts whatever data is sent to, or received from, the lower-level networking functions; at each successively lower layer, additional traffic is involved (such as header overhead, retransmits and fragmentation, unrelated broadcasts/traffic, etc.). The additional traffic is never seen by Squid and thus isn't counted - but if you run MRTG (or any SNMP/RMON measurement tool) against a specific interface, all this additional traffic will "magically appear".

Also remember that an interface has no concept of upper-layer networking (so an Ethernet interface doesn't distinguish between IP traffic that's entirely internal to your organization, and traffic that's to/from the Internet); this means that when you start measuring an interface, you have to be aware of *what* you are measuring before you can start comparing numbers elsewhere.

It is possible (though by no means guaranteed) that you are seeing roughly equivalent input/output because you're measuring an interface that both retrieves data from the outside world (Internet), *and* serves it to end users (internal clients). That wouldn't be the whole answer, but hopefully it gives you a few ideas to start applying to your own circumstance.

To interpret any statistic, you have to first know what you are measuring; for example, an interface counts inbound and outbound bytes - that's it. The interface doesn't distinguish between inbound bytes from external Internet sites or from internal (to the organization) clients (making requests). If you want that, try looking at RMON2.

Also, if you're talking about a 40% hit rate in terms of object requests/counts then there's absolutely no reason why you should expect a 40% reduction in traffic; after all, not every request/object is going to be the same size so you may be saving a lot in terms of requests but very little in terms of actual traffic. How come some objects do not get cached?

To determine whether a given object may be cached, Squid takes many things into consideration. The current algorithm (for Squid-2) goes something like this: Responses with Responses with Responses with Responses for requests with an Responses with The following HTTP status codes are cachable: 200 OK 203 Non-Authoritative Information 300 Multiple Choices 301 Moved Permanently 410 Gone However, if Squid receives one of these responses from a neighbor cache, it will NOT be cached if ALL of the A 302 Moved Temporarily response is cachable ONLY if the response also includes an The following HTTP status codes are ``negatively cached'' for a short amount of time (configurable): 204 No Content 305 Use Proxy 400 Bad Request 403 Forbidden 404 Not Found 405 Method Not Allowed 414 Request-URI Too Large 500 Internal Server Error 501 Not Implemented 502 Bad Gateway 503 Service Unavailable 504 Gateway Time-out All other HTTP status codes are NOT cachable, including: 206 Partial Content 303 See Other 304 Not Modified 401 Unauthorized 407 Proxy Authentication Required What does The This is a mechanism to try detecting neighbor caches which might not be able to deal with persistent connections. Every time we send a If the ratio stays above 0.5, then we continue to assume the neighbor properly implements persistent connections. Otherwise, we will stop sending the keep-alive request header to that neighbor. How does Squid's cache replacement algorithm work?

Squid uses an LRU (least recently used) algorithm to replace old cache objects. This means objects which have not been accessed for the longest time are removed first. In the source code, the StoreEntry->lastref value is updated every time an object is accessed.

Objects are not necessarily removed ``on-demand.'' Instead, a regularly scheduled event runs to periodically remove objects. Normally this event runs every second.

Squid keeps the cache disk usage between the low and high water marks. By default the low mark is 90%, and the high mark is 95% of the total configured cache size. When the disk usage is close to the low mark, the replacement is less aggressive (fewer objects removed). When the usage is close to the high mark, the replacement is more aggressive (more objects removed).

When selecting objects for removal, Squid examines some number of objects and determines which can be removed and which cannot. A number of factors determine whether or not any given object can be removed. If the object is currently being requested, or retrieved from an upstream site, it will not be removed. If the object is ``negatively-cached'' it will be removed. If the object has a private cache key, it will be removed (there would be no reason to keep it -- because the key is private, it can never be ``found'' by subsequent requests). Finally, if the time since last access is greater than the LRU threshold, the object is removed.

The LRU threshold value is dynamically calculated based on the current cache size and the low and high marks. The LRU threshold scaled exponentially between the high and low water marks. When the store swap size is near the low water mark, the LRU threshold is large. When the store swap size is near the high water mark, the LRU threshold is small. The threshold automatically adjusts to the rate of incoming requests. In fact, when your cache size has stabilized, the LRU threshold represents how long it takes to fill (or fully replace) your cache at the current request rate. Typical values for the LRU threshold are 1 to 10 days.

Back to selecting objects for removal. Obviously it is not possible to check every object in the cache every time we need to remove some of them. We can only check a small subset each time. The way in which this is implemented is very different between Squid-1.1 and Squid-2. Squid 1.1

The Squid cache storage is implemented as a hash table with some number of "hash buckets." Squid-1.1 scans one bucket at a time and sorts all the objects in the bucket by their LRU age. Objects with an LRU age over the threshold are removed. The scan rate is adjusted so that it takes approximately 24 hours to scan the entire cache. The store buckets are randomized so that we don't always scan the same buckets at the same time of the day.

This algorithm has some flaws. Because we only scan one bucket, there are going to be better candidates for removal in some of the other 16,000 or so buckets. Also, the qsort() function might take a non-trivial amount of CPU time, depending on how many entries are in each bucket. Squid 2

For Squid-2 we eliminated the need to use qsort() by indexing cached objects into an automatically sorted linked list. Every time an object is accessed, it gets moved to the top of the list. Over time, the least used objects migrate to the bottom of the list. When looking for objects to remove, we only need to check the last 100 or so objects in the list. Unfortunately this approach increases our memory usage because of the need to store three additional pointers per cache object. But for Squid-2 we're still ahead of the game because we also replaced plain-text cache keys with MD5 hashes. What are private and public keys?

The Squid cache uses the notions of Objects are changed from private to public after all of the HTTP reply headers have been received and parsed. In some cases, the reply headers will indicate the object should not be made public. For example, if the What is FORW_VIA_DB for?

We use it to collect data for . Does Squid send packets to port 7 (echo)? If so, why?

It may. This is an old feature from the Harvest cache software. The cache would send ICP ``SECHO'' message to the echo ports of origin servers. If the SECHO message came back before any of the other ICP replies, then it meant the origin server was probably closer than any neighbor cache. In that case Harvest/Squid sent the request directly to the origin server.

With more attention focused on security, many administrators filter UDP packets to port 7. The Computer Emergency Response Team (CERT) once issued an advisory note () that says UDP echo and chargen services can be used for a denial of service attack. This made admins extremely nervous about any packets hitting port 7 on their systems, and they made complaints.

The What does ``WARNING: Reply from unknown nameserver [a.b.c.d]'' mean?

It means Squid sent a DNS query to one IP address, but the response came back from a different IP address. By default Squid checks that the addresses match. If not, Squid ignores the response.

There are a number of reasons why this would happen: Your DNS name server just works this way, either becuase its been configured to, or because its stupid and doesn't know any better. You have a weird broadcast address, like, in your /etc/resolv.conf file. Somebody is trying to send spoofed DNS responses to your cache.

If you recognize the IP address in the warning as one of your name server hosts, then its probably numbers (1) or (2).

You can make these warnings stop, and allow responses from ``unknown'' name servers by setting this configuration option: ignore_unknown_nameservers off How does Squid distribute cache files among the available directories?

Note: The information here is current for version 2.2.

See When Squid wants to create a new disk file for storing an object, it first selects which 3N/4 (75%) of them with the most available space. These directories are then used, in order of having the most available space. When Squid has stored one URL to each of the 3N/4 3N/4 cache directories with the most available space.

Once the Why do I see negative byte hit ratio?

Byte hit ratio is calculated a bit differently than Request hit ratio. Squid counts the number of bytes read from the network on the server-side, and the number of bytes written to the client-side. The byte hit ratio is calculated as (client_bytes - server_bytes) / client_bytes If server_bytes is greater than client_bytes, you end up with a negative value.

The server_bytes may be greater than client_bytes for a number of reasons, including: Cache Digests and other internally generated requests. Cache Digest messages are quite large. They are counted in the server_bytes, but since they are consumed internally, they do not count in client_bytes. User-aborted requests. If your Some range requests, in combination with Squid bugs, can consume more bandwidth on the server-side than on the client-side. In a range request, the client is asking for only some part of the object. Squid may decide to retrieve the whole object anyway, so that it can be used later on. This means downloading more from the server than sending to the client. You can affect this behavior with the What does ``Disabling use of private keys'' mean?

First you need to understand the .

When Squid sends ICP queries, it uses the ICP Some ICP implementations always set the Not having private cache keys has some important privacy implications. Two users could receive one response that was meant for only one of the users. This response could contain personal, confidential information. You will need to disable the ``zero reqnum'' neighbor if you want Squid to use private cache keys. What is a half-closed filedescriptor?

TCP allows connections to be in a ``half-closed'' state. This is accomplished with the If Squid tries to read a connection, and To disable half-closed connections, simply put this in squid.conf: half_closed_clients off Then, Squid will always close its side of the connection instead of marking it as half-closed. What does --enable-heap-replacement do?

Squid has traditionally used an LRU replacement algorithm. As of , you can use some other replacement algorithms by using the With Squid version 2.4 and later you should use this configure option: ./configure --enable-removal-policies=heap

Then, in The LFUDA and GDS replacement code was contributed by John Dilley and others from Hewlett-Packard. Their work is described in these papers: (HP Tech Report). (WCW 1999 paper). Why is actual filesystem space used greater than what Squid thinks?

If you compare Squid doesn't keep track of the size of the Directory entries and take up filesystem space. Other applications might be using the same disk partition. Your filesystem block size might be larger than what Squid thinks. When calculating total disk usage, Squid rounds file sizes up to a whole number of 1024 byte blocks. If your filesystem uses larger blocks, then some "wasted" space is not accounted. How do Squid-2.3 and later versions with internal DNS lookups. Internal lookups are the default for Squid-2.3 and later. If you applied the ``DNS TTL'' for BIND. If you are using FreeBSD, then it already has the DNS TTL patch built in.

Let's say you have the following settings: positive_dns_ttl 1 hours negative_dns_ttl 1 minutes When Squid looks up a name like If you have the DNS TTL patch, or are using internal lookups, then each hostname has its own TTL value, which was set by the domain name administrator. You can see these values in the 'ipcache' cache manager page. For example: Hostname Flags lstref TTL N www.squid-cache.org C 73043 12784 1( 0) www.ircache.net C 73812 10891 1( 0) polygraph.ircache.net C 241768 -181261 1( 0) The TTL field shows how how many seconds until the entry expires. Negative values mean the entry is already expired, and will be refreshed upon next use.

The What does swapin MD5 mismatch mean?

It means that Squid opened up a disk file to serve a cache hit, but it found that the stored object doesn't match what the user's request. Squid stores the MD5 digest of the URL at the start of each disk file. When the file is opened, Squid checks that the disk file MD5 matches the MD5 of the URL requested by the user. If they don't match, the warning is printed and Squid forwards the request to the origin server.

You do not need to worry about this warning. It means that Squid is recovering from a corrupted cache directory. What does failed to unpack swapfile meta data mean?

Each of Squid's disk cache files has a metadata section at the beginning. This header is used to store the URL MD5, some StoreEntry data, and more. When Squid opens a disk file for reading, it looks for the meta data header and unpacks it.

This warning means that Squid couln't unpack the meta data. This is non-fatal bug, from which Squid can recover. Perhaps the meta data was just missing, or perhaps the file got corrupted.

You do not need to worry about this warning. It means that Squid is double-checking that the disk file matches what Squid thinks should be there, and the check failed. Squid recorvers and generates a cache miss in this case. Why doesn't Squid make Its a side-effect of the way interception proxying works.

When Squid is configured for interception proxying, the operating system pretends that it is the origin server. That means that the "local" socket address for intercepted TCP connections is really the origin server's IP address. If you run When Squid wants to make an ident query, it creates a new TCP socket and So why bind in that way? If you know you are interception proxying, then why not bind the local endpoint to the host's (intranet) IP address? Why make the masses suffer needlessly?

Because thats just how ident works. Please read , in particular the RESTRICTIONS section. dnsSubmit: queue overload, rejecting blah

This means that you are using external To alleviate this condition, you need to either (1) increase the number of Note that in some versions, Squid limits What are FTP passive connections?

by Colin Campbell

Ftp uses two data streams, one for passing commands around, the other for moving data. The command channel is handled by the ftpd listening on port 21.

The data channel varies depending on whether you ask for passive ftp or not. When you request data in a non-passive environment, you client tells the server ``I am listening on <ip-address> <port>.'' The server then connects FROM port 20 to the ip address and port specified by your client. This requires your "security device" to permit any host outside from port 20 to any host inside on any port > 1023. Somewhat of a hole.

In passive mode, when you request a data transfer, the server tells the client ``I am listening on <ip address> <port>.'' Your client then connects to the server on that IP and port and data flows. Multicast What is Multicast?

Multicast is essentially the ability to send one IP packet to multiple receivers. Multicast is often used for audio and video conferencing systems. How do I know if my network has multicast?

One way is to ask someone who manages your network. If your network manager doesn't know, or looks at you funny, then you probably don't have it.

Another way is to use the . Mtrace is similar to traceroute. It will tell you about the multicast path between your site and another. For example: > mtrace mbone.ucar.edu mtrace: WARNING: no multicast group specified, so no statistics printed Mtrace from to via group Querying full reverse path... * switching to hop-by-hop: 0 oceana-ether.nlanr.net ( -1 avidya-ether.nlanr.net ( DVMRP thresh^ 1 -2 mbone.sdsc.edu ( DVMRP thresh^ 1 -3 * nccosc-mbone.dren.net ( DVMRP thresh^ 48 -4 * * FIXW-MBONE.NSN.NASA.GOV ( PIM/Special thresh^ 64 -5 dec3800-2-fddi-0.SanFrancisco.mci.net ( DVMRP thresh^ 64 -6 dec3800-2-fddi-0.Denver.mci.net ( DVMRP thresh^ 1 -7 mbone.ucar.edu ( DVMRP thresh^ 64 -8 mbone.ucar.edu ( Round trip time 196 ms; total ttl of 68 required. Should I be using Multicast ICP?

Short answer: No, probably not.

Reasons why you SHOULD use Multicast: It reduces the number of times Squid calls Its trendy and cool to use Multicast.

Reasons why you SHOULD NOT use Multicast: Multicast tunnels/configurations/infrastructure are often unstable. You may lose multicast connectivity but still have unicast connectivity. Multicast does not simplify your Squid configuration file. Every trusted neighbor cache must still be specified. Multicast does not reduce the number of ICP replies being sent around. It does reduce the number of ICP queries sent, but not the number of replies. Multicast exposes your cache to some privacy issues. There are no special emissions required to join a multicast group. Anyone may join your group and eavesdrop on ICP query messages. However, the scope of your multicast traffic can be controlled such that it does not exceed certain boundaries.

We only recommend people to use Multicast ICP over network infrastructure which they have close control over. In other words, only use Multicast over your local area network, or maybe your wide area network if you are an ISP. We think it is probably a bad idea to use Multicast ICP over congested links or commodity backbones. How do I configure Squid to send Multicast ICP queries?

To configure Squid to send ICP queries to a Multicast address, you need to create another neighbour cache entry specified as cache_peer multicast 3128 3130 ttl=64 is a sample multicast group address. You must also specify which of your neighbours will respond to your multicast queries, since it would be a bad idea to implicitly trust any ICP reply from an unknown address. Note that ICP replies are sent back to cache_peer cache1 sibling 3128 3130 multicast-responder cache_peer cache2 sibling 3128 3130 multicast-responder Here all fields are relevant. The ICP port number (3130) must be the same as in the How do I know what Multicast TTL to use?

The Multicast TTL (which is specified on the 32 for links that separate sites within an organization. 64 for links that separate communities or organizations, and are attached to the Internet MBONE. 128 for links that separate continents on the MBONE.

A good way to determine the TTL you need is to run If you set you TTL too high, then your ICP messages may travel ``too far'' and will be subject to eavesdropping by others. If you're only using multicast on your LAN, as we suggest, then your TTL will be quite small, for example How do I configure Squid to receive and respond to Multicast ICP?

You must tell Squid to join a multicast group address with the mcast_groups Of course, all members of your Multicast ICP group will need to use the exact same multicast group address.

Use a unique group address Limit the scope of multicast messages with TTLs or administrative scoping.

Using a unique address is a good idea, but not without some potential problems. If you choose an address randomly, how do you know that someone else will not also randomly choose the same address? NLANR has been assigned a block of multicast addresses by the IANA for use in situations such as this. If you would like to be assigned one of these addresses, please . However, note that NLANR or IANA have no authority to prevent anyone from using an address assigned to you.

Limiting the scope of your multicast messages is probably a better solution. They can be limited with the TTL value discussed above, or with some newer techniques known as administratively scoped addresses. Here you can configure well-defined boundaries for the traffic to a specific address. The describes this. System-Dependent Weirdnesses Solaris TCP incompatibility?

J.D. Bronson (jb at ktxg dot com) reported that his Solaris box could not talk to certain origin servers, such as and . J.D. fixed his problem by setting: tcp_xmit_hiwat 49152 tcp_xmit_lowat 4096 tcp_recv_hiwat 49152 select()

For older Squid versions you can enable include/autoconf.h, or by adding -DUSE_POLL=1 to the DEFINES in src/Makefile. malloc

libmalloc.a is leaky. Squid's configure does not use -lmalloc on Solaris. DNS lookups and by .

DNS lookups can be slow because of some mysterious thing called /etc/nscd.conf and make it say: enable-cache hosts no

Apparently nscd serializes DNS queries thus slowing everything down when an application (such as Squid) hits the resolver hard. You may notice something similar if you run a log processor executing many DNS resolver queries - the resolver starts to slow.. right.. down.. . . .

According to , users of Solaris starting from version 2.6 and up should NOT completely disable Several library calls rely on available free FILE descriptors FD < 256. Systems running without nscd may fail on such calls if first 256 files are all in use.

Since solaris 2.6 Sun has changed the way some system calls work and is using DNS lookups and /etc/nsswitch.conf

by .

The /etc/nsswitch.conf file determines the order of searches for lookups (amongst other things). You might only have it set up to allow NIS and HOSTS files to work. You definitely want the "hosts:" line to include the word hosts: nis dns [NOTFOUND=return] files DNS lookups and NIS

by .

Our site cache is running on a Solaris 2.6 machine. We use NIS to distribute authentication and local hosts information around and in common with our multiuser systems, we run a slave NIS server on it to help the response of NIS queries.

We were seeing very high name-ip lookup times (avg ˜2sec) and ip->name lookup times (avg ˜8 sec), although there didn't seem to be that much of a problem with response times for valid sites until the cache was being placed under high load. Then, performance went down the toilet.

After some time, and a bit of detective work, we found the problem. On Solaris 2.6, if you have a local NIS server running (/etc/nsswitch.conf hosts entry, then check the flags it is being started with. The 2.6 ypstart script checks to see if there is a This has the same effect as putting the This is a If you're running in this kind of setup, then you will want to make sure that ypserv doesn't start with the you don't have the

We changed these here, and saw our average lookup times drop by up to an order of magnitude (˜150msec for name-ip queries and ˜1.5sec for ip-name queries, the latter still so high, I suspect, because more of these fail and timeout since they are not made so often and the entries are frequently non-existent anyway). Tuning

by disk write error: (28) No space left on device

You might get this error even if your disk is not full, and is not out of inodes. Check your syslog logs (/var/adm/messages, normally) for messages like either of these: NOTICE: realloccg /proxy/cache: file system full NOTICE: alloc: /proxy/cache: file system full

In a nutshell, the UFS filesystem used by Solaris can't cope with the workload squid presents to it very well. The filesystem will end up becoming highly fragmented, until it reaches a point where there are insufficient free blocks left to create files with, and only fragments available. At this point, you'll get this error and squid will revise its idea of how much space is actually available to it. You can do a "fsck -n raw_device" (no need to unmount, this checks in read only mode) to look at the fragmentation level of the filesystem. It will probably be quite high (>15%).

Sun suggest two solutions to this problem. One costs money, the other is free but may result in a loss of performance (although Sun do claim it shouldn't, given the already highly random nature of squid disk access).

The first is to buy a copy of VxFS, the Veritas Filesystem. This is an extent-based filesystem and it's capable of having online defragmentation performed on mounted filesystems. This costs money, however (VxFS is not very cheap!)

The second is to change certain parameters of the UFS filesystem. Unmount your cache filesystems and use tunefs to change optimization to "space" and to reduce the "minfree" value to 3-5% (under Solaris 2.6 and higher, very large filesystems will almost certainly have a minfree of 2% already and you shouldn't increase this). You should be able to get fragmentation down to around 3% by doing this, with an accompanied increase in the amount of space available.

Thanks to . Solaris X86 and IPFilter


Important update regarding Squid running on Solaris x86. I have been working for several months to resolve what appeared to be a memory leak in squid when running on Solaris x86 regardless of the malloc that was used. I have made 2 discoveries that anyone running Squid on this platform may be interested in.

Number 1: There is not a memory leak in Squid even though after the system runs for some amount of time, this varies depending on the load the system is under, Top reports that there is very little memory free. True to the claims of the Sun engineer I spoke to this statistic from Top is incorrect. The odd thing is that you do begin to see performance suffer substantially as time goes on and the only way to correct the situation is to reboot the system. This leads me to discovery number 2.

Number 2: There is some type of resource problem, memory or other, with IPFilter on Solaris x86. I have not taken the time to investigate what the problem is because we no longer are using IPFilter. We have switched to a Alteon ACE 180 Gigabit switch which will do the trans-proxy for you. After moving the trans-proxy, redirection process out to the Alteon switch Squid has run for 3 days strait under a huge load with no problem what so ever. We currently have 2 boxes with 40 GB of cached objects on each box. This 40 GB was accumulated in the 3 days, from this you can see what type of load these boxes are under. Prior to this change we were never able to operate for more than 4 hours.

Because the problem appears to be with IPFilter I would guess that you would only run into this issue if you are trying to run Squid as a interception proxy using IPFilter. That makes sense. If there is anyone with information that would indicate my finding are incorrect I am willing to investigate further. Changing the directory lookup cache size


On Solaris, the kernel variable for the directory name lookup cache size is ncsize. In /etc/system, you might want to try set ncsize = 8192 or even higher. The kernel variable You can set Defaults are: Solaris 2.5.1 : (max_nprocs + 16 + maxusers) + 64 Solaris 2.6/Solaris 7 : 4 * (max_nprocs + maxusers) + 320 The priority_paging algorithm


Another new tuneable (actually a toggle) in Solaris 2.5.1, 2.6 or Solaris 7 is the /etc/system). As you may know, the Solaris buffer cache grows to fill available pages, and under the old VM system, applications could get paged out to make way for the buffer cache, which can lead to swap thrashing and degraded application performance. The new FreeBSD T/TCP bugs

We have found that with FreeBSD-2.2.2-RELEASE, there some bugs with T/TCP. FreeBSD will try to use T/TCP if you've enabled the ``TCP Extensions.'' To disable T/TCP, use /etc/rc.conf and set tcp_extensions="NO" # Allow RFC1323 & RFC1544 extensions (or NO). or add this to your /etc/rc files: sysctl -w net.inet.tcp.rfc1644=0 mbuf size

We noticed an odd thing with some of Squid's interprocess communication. Often, output from the 1998/04/02 15:18:48| comm_select: FD 46 ready for reading 1998/04/02 15:18:48| ipcache_dnsHandleRead: Result from DNS ID 2 (100 bytes) 1998/04/02 15:18:48| ipcache_dnsHandleRead: Incomplete reply ....other processing occurs... 1998/04/02 15:18:48| comm_select: FD 46 ready for reading 1998/04/02 15:18:48| ipcache_dnsHandleRead: Result from DNS ID 2 (9 bytes) 1998/04/02 15:18:48| ipcache_parsebuffer: parsing: $name www.karup.com $h_name www.karup.inter.net $h_len 4 $ipcount 2 $ttl 2348 $end Interestingly, it is very common to get only 100 bytes on the first read. When two read() calls are required, this adds additional latency to the overall request. On our caches running Digital Unix, the median Here is a simple patch to fix the bug: =================================================================== RCS file: /home/ncvs/src/sys/kern/uipc_socket.c,v retrieving revision 1.40 retrieving revision 1.41 diff -p -u -r1.40 -r1.41 --- src/sys/kern/uipc_socket.c 1998/05/15 20:11:30 1.40 +++ /home/ncvs/src/sys/kern/uipc_socket.c 1998/07/06 19:27:14 1.41 @@ -31,7 +31,7 @@ * SUCH DAMAGE. * * @(#)uipc_socket.c 8.3 (Berkeley) 4/15/94 - * $Id: FAQ.sgml,v 1.215 2004/02/05 17:06:11 wessels Exp $ + * $Id: FAQ.sgml,v 1.215 2004/02/05 17:06:11 wessels Exp $ */ #include @@ -491,6 +491,7 @@ restart: mlen = MCLBYTES; len = min(min(mlen, resid), space); } else { + atomic = 1; nopages: len = min(min(mlen, resid), space); /*

Another technique which may help, but does not fix the bug, is to increase the kernel's mbuf size. The default is 128 bytes. The MSIZE symbol is defined in /usr/include/machine/param.h. However, to change it we added this line to our kernel configuration file: options MSIZE="256" Dealing with NIS

/var/yp/Makefile has the following section: # The following line encodes the YP_INTERDOMAIN key into the hosts.byname # and hosts.byaddr maps so that ypserv(8) will do DNS lookups to resolve # hosts not in the current domain. Commenting this line out will disable # the DNS lookups. B=-b You will want to comment out the FreeBSD 3.3: The lo0 (loop-back) device is not configured on startup

Squid requires a the loopback interface to be up and configured. If it is not, you will get errors such as .

From :

Fix: Assuming that you experience this problem at all, edit /etc/rc.conf and search for where the network_interfaces variable is set. In its value, change the word

Thanks to . FreeBSD 3.x or newer: Speed up disk writes using Softupdates


FreeBSD 3.x and newer support Softupdates. This is a mechanism to speed up disk writes as it is possible by mounting ufs volumes async. However, Softupdates does this in a way that a performance similar or better than async is achieved but without loosing security in a case of a system crash. For more detailed information and the copyright terms see /sys/contrib/softupdates/README and /sys/ufs/ffs/README.softupdate.

To build a system supporting softupdates, you have to build a kernel with options SOFTUPDATES set (see $ tunefs -n /mountpoint The filesystem in question MUST NOT be mounted at this time. After that, softupdates are permanently enabled and the filesystem can be mounted normally. To verify that the softupdates code is running, simply issue a mount command and an output similar to the following will appear: $ mount /dev/da2a on /usr/local/squid/cache (ufs, local, noatime, soft-updates, writes: sync 70 async 225) Internal DNS problems with jail environment

Some users report problems with running Squid in the jail environment. Specifically, Squid logs messages like: 2001/10/12 02:08:49| comm_udp_sendto: FD 4,, port 53: (22) Invalid argument 2001/10/12 02:08:49| idnsSendQuery: FD 4: sendto: (22) Invalid argument

You can eliminate the problem by putting the jail's network interface address in the 'udp_outgoing_addr' configuration option in squid.conf. OSF1/3.2

If you compile both libgnumalloc.a and Squid with BSD/OS gcc/yacc

Some people report . process priority

I've noticed that my Squid process seems to stick at a nice value of four, and clicks back to that even after I renice it to a higher priority. However, looking through the Squid source, I can't find any instance of a setpriority() call, or anything else that would seem to indicate Squid's adjusting its own priority.


BSD Unices traditionally have auto-niced non-root processes to 4 after they used alot (4 minutes???) of CPU time. My guess is that it's the BSD/OS not Squid that is doing this. I don't know offhand if there is a way to disable this on BSD/OS.


You can get around this by starting Squid with nice-level -4 (or another negative value).


The autonice behavior is a leftover from the history of BSD as a university OS. It penalises CPU bound jobs by nicing them after using 600 CPU seconds. Adding sysctl -w kern.autonicetime=0 to /etc/rc.local will disable the behavior systemwide. Linux Cannot bind socket FD 5 to (49) Can't assign requested address

Try a different version of Linux. We have received many reports of this ``bug'' from people running Linux 2.0.30. The FATAL: Don't run Squid as root, set 'cache_effective_user'!

Some users have reported that setting Another problem is that RedHat 5.0 Linux seems to have a broken % setenv ac_cv_func_setresuid no % ./configure ... % make clean % make install Or after running configure, manually edit include/autoconf.h and change the HAVE_SETRESUID line to: #define HAVE_SETRESUID 0

Also, some users report this error is due to a NIS configuration problem. By adding /etc/nsswitch.conf, the problem goes away. ().

notes that these problems with Large ACL lists make Squid slow

The regular expression library which comes with Linux is known to be very slow. Some people report it entirely fails to work after long periods of time.

To fix, use the GNUregex library included with the Squid source code. With Squid-2, use the gethostbyname() leaks memory in RedHat 6.0 with glibc 2.1.1.


The gethostbyname() function leaks memory in RedHat 6.0 with glibc 2.1.1. The quick fix is to delete nisplus service from hosts entry in /etc/nsswitch.conf. In my tests dnsserver memory use remained stable after I made the above change.

See . assertion failed: StatHist.c:91: `statHistBin(H, max) == H->capacity - 1' on Alpha system.


Some early versions of Linux have a kernel bug that causes this. All that is needed is a recent kernel that doesn't have the mentioned bug. tools.c:605: storage size of `rl' isn't known

This is a bug with some versions of glibc. The glibc headers incorrectly depended on the contents of some kernel headers. Everything broke down when the kernel folks rearranged a bit in the kernel-specific header files.

We think this glibc bug is present in versions 2.1.1 (or 2.1.0) and earlier. There are two solutions: Make sure /usr/include/linux and /usr/include/asm are from the kernel version glibc is build/configured for, not any other kernel version. Only compiling of loadable kernel modules outside of the kernel sources depends on having the current versions of these, and for such builds -I/usr/src/linux/include (or where ever the new kernel headers are located) can be used to resolve the matter. Upgrade glibc to 2.1.2 or later. This is always a good idea anyway, provided a prebuilt upgrade package exists for the Linux distribution used.. Note: Do not attempt to manually build and install glibc from source unless you know exactly what you are doing, as this can easily render the system unuseable. Can't connect to some sites through Squid

When using Squid, some sites may give erorrs such as ``(111) Connection refused'' or ``(110) Connection timed out'' although these sites work fine without going through Squid.

Some versions of linux implement (ECN) and this can cause some TCP connections to fail when contacting some sites with broken firewalls or broken TCP/IP implementations. A list of sites to be broken can be found at . To work around such broken sites you can disable ECN with the following command: echo 0 > /proc/sys/net/ipv4/tcp_ecn

Found this on the FreeBSD mailing list:

From: Robert Watson

As Bill Fumerola has indicated, and I thought I'd follow up in with a bit more detail, the behavior you're seeing is the result of a bug in the FreeBSD IPFW code. FreeBSD did a direct comparison of the TCP header flag field with an internal field in the IPFW rule description structure. Unfortunately, at some point, someone decided to overload the IPFW rule description structure field to add a flag representing "ESTABLISHED". They used a flag value that was previously unused by the TCP protocol (which doesn't make it safer, just less noticeable). Later, when that flag was allocated for ECN (Endpoint Congestion Notification) in TCP, and Linux began using ECN by default, the packets began to match ESTABLISHED rules regardless of the other TCP header flags. This bug was corrected on the RELENG_4 branch, and security advisory for the bug was released. This was, needless to say, a pretty serious bug, and good example of why you should be very careful to compare only the bits you really mean to, and should seperate packet state from protocol state in management structures, as well as make use of extensive testing to make sure rules actually have the effect you describe.

See also the , , or for more information. HP-UX StatHist.c:74: failed assertion `statHistBin(H, min) == 0'

This was a very mysterious and unexplainable bug with GCC on HP-UX. Certain functions, when specified as IRIX There is a problem with GCC (2.8.1 at least) on Irix 6 which causes it to always return the string for _ANY_ address when calling inet_ntoa(). If this happens to you, compile Squid with the native C compiler instead of GCC. SCO-UNIX


To make squid run comfortable on SCO-unix you need to do the following:

Increase the One thing left is the number of tcp-connections the system can handle. Default is 256, but I increase that as well because of the number of clients we have. AIX "shmat failed" errors with 32-bit processes on AIX and later are restricted by default to a maximum of 11 shared memory segments. This restriction can be removed on AIX 4.2.1 and later by setting the environment variable Core dumps when squid process grows to 256MB

32-bit processes cannot use more than 256MB of stack and data in the default memory model. To force the loader to use large address space for squid, either: set the link with patch the squid binary See on large program support for more information, including how to patch an already-compiled program. Redirectors What is a redirector?

Squid has the ability to rewrite requested URLs. Implemented as an external process (similar to a dnsserver), Squid can be configured to pass every incoming URL through a The Why use a redirector?

A redirector allows the administrator to control the locations to which his users goto. Using this in conjunction with interception proxies allows simple but effective porn control. How does it work?

The redirector program must read URLs (one per line) on standard input, and write rewritten URLs or blank lines on standard output. Note that the redirector program can not use buffered I/O. Squid writes additional information after the URL which a redirector can use to make a decision. The input line consists of four fields: URL ip-address/fqdn ident method Do you have any examples?

A simple very fast redirector called is a good place to start, it uses the regex lib to allow pattern matching.

Also see .

The following Perl script may also be used as a template for writing your own redirector: #!/usr/local/bin/perl $|=1; while (<>) { s@http://fromhost.com@http://tohost.org@; print; } Can I use the redirector to return HTTP redirect messages?

Normally, the Simply modify your redirector program to prepend either "301:" or "302:" before the new URL. For example, the following script might be used to direct external clients to a secure Web server for internal documents: #!/usr/local/bin/perl $|=1; while (<>) { @X = split; $url = $X[0]; if ($url =~ /^http:\/\/internal\.foo\.com/) { $url =~ s/^http/https/; $url =~ s/internal/secure/; print "302:$url\n"; } else { print "$url\n"; } }

Please see sections 10.3.2 and 10.3.3 of for an explanation of the 301 and 302 HTTP reply codes. FATAL: All redirectors have exited!

A redirector process must exit (stop running) only when its files from the redirector program. Redirector interface is broken re IDENT values

I added a redirctor consisting of #! /bin/sh /usr/bin/tee /tmp/squid.log and many of the redirector requests don't have a username in the ident field.

Squid does not delay a request to wait for an ident lookup, unless you use the ident ACLs. Thus, it is very likely that the ident was not available at the time of calling the redirector, but became available by the time the request is complete and logged to access.log.

If you want to block requests waiting for ident lookup, try something like this: acl foo ident REQUIRED http_access allow foo Cache Digests

Cache Digest FAQs compiled by . What is a Cache Digest?

A Cache Digest is a summary of the contents of an Internet Object Caching Server. It contains, in a compact (i.e. compressed) format, an indication of whether or not particular URLs are in the cache.

A "lossy" technique is used for compression, which means that very high compression factors can be achieved at the expense of not having 100% correct information. How and why are they used?

Cache servers periodically exchange their digests with each other.

When a request for an object (URL) is received from a client a cache can use digests from its peers to find out which of its peers (if any) have that object. The cache can then request the object from the closest peer (Squid uses the NetDB database to determine this).

Note that Squid will only make digest queries in those digests that are enabled. It will disable a peers digest IFF it cannot fetch a valid digest for that peer. It will enable that peers digest again when a valid one is fetched.

The checks in the digest are very fast and they eliminate the need for per-request queries to peers. Hence:

Latency is eliminated and client response time should be improved. Network utilisation may be improved.

Note that the use of Cache Digests (for querying the cache contents of peers) and the generation of a Cache Digest (for retrieval by peers) are independent. So, it is possible for a cache to make a digest available for peers, and not use the functionality itself and vice versa. What is the theory behind Cache Digests?

Cache Digests are based on Bloom Filters - they are a method for representing a set of keys with lookup capabilities; where lookup means "is the key in the filter or not?".

In building a cache digest:

A vector (1-dimensional array) of m bits is allocated, with all bits initially set to 0. A number, k, of independent hash functions are chosen, h1, h2, ..., hk, with range { 1, ..., m } (i.e. a key hashed with any of these functions gives a value between 1 and m inclusive). The set of n keys to be operated on are denoted by: A = { a1, a2, a3, ..., an }. Adding a Key

To add a key the value of each hash function for that key is calculated. So, if the key was denoted by a, then h1(a), h2(a), ..., hk(a) are calculated.

The value of each hash function for that key represents an index into the array and the corresponding bits are set to 1. So, a digest with 6 hash functions would have 6 bits to be set to 1 for each key added.

Note that the addition of a number of different keys could cause one particular bit to be set to 1 multiple times. Querying a Key

To query for the existence of a key the indices into the array are calculated from the hash functions as above.

If any of the corresponding bits in the array are 0 then the key is not present. If all of the corresponding bits in the array are 1 then the key is likely to be present.

Note the term likely. It is possible that a collision in the digest can occur, whereby the digest incorrectly indicates a key is present. This is the price paid for the compact representation. While the probability of a collision can never be reduced to zero it can be controlled. Larger values for the ratio of the digest size to the number of entries added lower the probability. The number of hash functions chosen also influence the probability. Deleting a Key

To delete a key, it is not possible to simply set the associated bits to 0 since any one of those bits could have been set to 1 by the addition of a different key!

Therefore, to support deletions a counter is required for each bit position in the array. The procedures to follow would be:

When adding a key, set appropriate bits to 1 and increment the corresponding counters. When deleting a key, decrement the appropriate counters (while > 0), and if a counter reaches 0 then the corresponding bit is set to 0. How is the size of the Cache Digest in Squid determined?

Upon initialisation, the capacity is set to the number of objects that can be (are) stored in the cache. Note that there are upper and lower limits here.

An arbitrary constant, bits_per_entry (currently set to 5), is used to calculate the size of the array using the following formula:

number of bits in array = capacity * bits_per_entry + 7

The size of the digest, in bytes, is therefore:

digest size = int (number of bits in array / 8)

When a digest rebuild occurs, the change in the cache size (capacity) is measured. If the capacity has changed by a large enough amount (10%) then the digest array is freed and reallocated memory, otherwise the same digest is re-used. What hash functions (and how many of them) does Squid use?

The protocol design allows for a variable number of hash functions (k). However, Squid employs a very efficient method using a fixed number - four.

Rather than computing a number of independent hash functions over a URL Squid uses a 128-bit MD5 hash of the key (actually a combination of the URL and the HTTP retrieval method) and then splits this into four equal chunks. Each chunk, modulo the digest size (m), is used as the value for one of the hash functions - i.e. an index into the bit array.

Note: As Squid retrieves objects and stores them in its cache on disk, it adds them to the in-RAM index using a lookup key which is an MD5 hash - the very one discussed above. This means that the values for the Cache Digest hash functions are already available and consequently the operations are extremely efficient!

Obviously, modifying the code to support a variable number of hash functions would prove a little more difficult and would most likely reduce efficiency. How are objects added to the Cache Digest in Squid?

Every object referenced in the index in RAM is checked to see if it is suitable for addition to the digest. A number of objects are not suitable, e.g. those that are private, not cachable, negatively cached etc. and are skipped immediately.

A freshness test is next made in an attempt to guess if the object will expire soon, since if it does, it is not worthwhile adding it to the digest. The object is checked against the refresh patterns for staleness...

Since Squid stores references to objects in its index using the MD5 key discussed earlier there is no URL actually available for each object - which means that the pattern used will fall back to the default pattern, ".". This is an unfortunate state of affairs, but little can be done about it. A cd_refresh_pattern option will be added to the configuration file soon which will at least make the confusion a little clearer :-)

Note that it is best to be conservative with your refresh pattern for the Cache Digest, i.e. do not add objects if they might become stale soon. This will reduce the number of False Hits. Does Squid support deletions in Cache Digests? What are diffs/deltas?

Squid does not support deletions from the digest. Because of this the digest must, periodically, be rebuilt from scratch to erase stale bits and prevent digest pollution.

A more sophisticated option is to use diffs or deltas. These would be created by building a new digest and comparing with the current/old one. They would essentially consist of aggregated deletions and additions since the previous digest.

Since less bandwidth should be required using these it would be possible to have more frequent updates (and hence, more accurate information).


RAM - extra RAM needed to hold two digests while comparisons takes place. CPU - probably a negligible amount. When and how often is the local digest built?

The local digest is built:

when store_rebuild completes after startup (the cache contents have been indexed in RAM), and periodically thereafter. Currently, it is rebuilt every hour (more data and experience is required before other periods, whether fixed or dynamically varying, can "intelligently" be chosen). The good thing is that the local cache decides on the expiry time and peers must obey (see later).

While the [new] digest is being built in RAM the old version (stored on disk) is still valid, and will be returned to any peer requesting it. When the digest has completed building it is then swapped out to disk, overwriting the old version.

The rebuild is CPU intensive, but not overly so. Since Squid is programmed using an event-handling model, the approach taken is to split the digest building task into chunks (i.e. chunks of entries to add) and to register each chunk as an event. If CPU load is overly high, it is possible to extend the build period - as long as it is finished before the next rebuild is due!

It may prove more efficient to implement the digest building as a separate process/thread in the future... How are Cache Digests transferred between peers?

Cache Digests are fetched from peers using the standard HTTP protocol (note that a pull rather than push technique is used).

After the first access to a peer, a peerDigestValidate event is queued (this event decides if it is time to fetch a new version of a digest from a peer). The queuing delay depends on the number of peers already queued for validation - so that all digests from different peers are not fetched simultaneously.

A peer answering a request for its digest will specify an expiry time for that digest by using the HTTP Expires header. The requesting cache thus knows when it should request a fresh copy of that peers digest.

Note: requesting caches use an If-Modified-Since request in case the peer has not rebuilt its digest for some reason since the last time it was fetched. How and where are Cache Digests stored?

Cache Digest built locally

Since the local digest is generated purely for the benefit of its neighbours keeping it in RAM is not strictly required. However, it was decided to keep the local digest in RAM partly because of the following:

Approximately the same amount of memory will be (re-)allocated on every rebuild of the digest, the memory requirements are probably quite small (when compared to other requirements of the cache server), if ongoing updates of the digest are to be supported (e.g. additions/deletions) it will be necessary to perform these operations on a digest in RAM, and if diffs/deltas are to be supported the "old" digest would have to be swapped into RAM anyway for the comparisons.

When the digest is built in RAM, it is then swapped out to disk, where it is stored as a "normal" cache item - which is how peers request it. Cache Digest fetched from peer

When a query from a client arrives, fast lookups are required to decide if a request should be made to a neighbour cache. It it therefore required to keep all peer digests in RAM.

Peer digests are also stored on disk for the following reasons:

Recovery - If stopped and restarted, peer digests can be reused from the local on-disk copy (they will soon be validated using an HTTP IMS request to the appropriate peers as discussed earlier), and Sharing - peer digests are stored as normal objects in the cache. This allows them to be given to neighbour caches. How are the Cache Digest statistics in the Cache Manager to be interpreted?

Cache Digest statistics can be seen from the Cache Manager or through the squidclient utility. The following examples show how to use the squidclient utility to request the list of possible operations from the localhost, local digest statistics from the localhost, refresh statistics from the localhost and local digest statistics from another cache, respectively.

squidclient mgr:menu squidclient mgr:store_digest squidclient mgr:refresh squidclient -h peer mgr:store_digest

The available statistics provide a lot of useful debugging information. The refresh statistics include a section for Cache Digests which explains why items were added (or not) to the digest.

The following example shows local digest statistics for a 16GB cache in a corporate intranet environment (may be a useful reference for the discussion below).

store digest: size: 768000 bytes entries: count: 588327 capacity: 1228800 util: 48% deletion attempts: 0 bits: per entry: 5 on: 1953311 capacity: 6144000 util: 32% bit-seq: count: 2664350 avg.len: 2.31 added: 588327 rejected: 528703 ( 47.33 %) del-ed: 0 collisions: on add: 0.23 % on rej: 0.23 %

entries:capacity is a measure of how many items "are likely" to be added to the digest. It represents the number of items that were in the local cache at the start of digest creation - however, upper and lower limits currently apply. This value is multiplied by bits: per entry (an arbitrary constant) to give bits:capacity, which is the size of the cache digest in bits. Dividing this by 8 will give store digest: size which is the size in bytes.

The number of items represented in the digest is given by entries:count. This should be equal to added minus deletion attempts. Since (currently) no modifications are made to the digest after the initial build (no additions are made and deletions are not supported) deletion attempts will always be 0 and entries:count should simply be equal to added.

entries:util is not really a significant statistic. At most it gives a measure of how many of the items in the store were deemed suitable for entry into the cache compared to how many were "prepared" for.

rej shows how many objects were rejected. Objects will not be added for a number of reasons, the most common being refresh pattern settings. Remember that (currently) the default refresh pattern will be used for checking for entry here and also note that changing this pattern can significantly affect the number of items added to the digest! Too relaxed and False Hits increase, too strict and False Misses increase. Remember also that at time of validation (on the peer) the "real" refresh pattern will be used - so it is wise to keep the default refresh pattern conservative.

bits: on indicates the number of bits in the digest that are set to 1. bits: util gives this figure as a percentage of the total number of bits in the digest. As we saw earlier, a figure of 50% represents the optimal trade-off. Values too high (say > 75%) would cause a larger number of collisions, and hence False Hits, while lower values mean the digest is under-utilised (using unnecessary RAM). Note that low values are normal for caches that are starting to fill up.

A bit sequence is an uninterrupted sequence of bits with the same value. bit-seq: avg.len gives some insight into the quality of the hash functions. Long values indicate problem, even if bits:util is 50% (> 3 = suspicious, > 10 = very suspicious). What are False Hits and how should they be handled?

A False Hit occurs when a cache believes a peer has an object and asks the peer for it but the peer is not able to satisfy the request.

Expiring or stale objects on the peer are frequent causes of False Hits. At the time of the query actual refresh patterns are used on the peer and stale entries are marked for revalidation. However, revalidation is prohibited unless the peer is behaving as a parent, or miss_access is enabled. Thus, clients can receive error messages instead of revalidated objects!

The frequency of False Hits can be reduced but never eliminated completely, therefore there must be a robust way of handling them when they occur. The philosophy behind the design of Squid is to use lightweight techniques and optimise for the common case and robustly handle the unusual case (False Hits).

Squid will soon support the HTTP only-if-cached header. Requests for objects made to a peer will use this header and if the objects are not available, the peer can reply appropriately allowing Squid to recognise the situation. The following describes what Squid is aiming towards:

Cache Digests used to obtain good estimates of where a requested object is located in a Cache Hierarchy. Persistent HTTP Connections between peers. There will be no TCP startup overhead and both latency and network load will be similar for ICP (i.e. fast). HTTP False Hit Recognition using the only-if-cached HTTP header - allowing fall back to another peer or, if no other peers are available with the object, then going direct (or through a parent if behind a firewall). How can Cache Digest related activity be traced/debugged?

Enabling Cache Digests

If you wish to use Cache Digests (available in Squid version 2) you need to add a configure option, so that the relevant code is compiled in:

./configure --enable-cache-digests ... What do the access.log entries look like?

If a request is forwarded to a neighbour due a HIT in that neighbour's Cache Digest the hierarchy (9th) field of the access.log file for the local cache will look like CACHE_DIGEST_HIT/neighbour. The Log Tag (4th field) should obviously show a MISS.

On the peer cache the request should appear as a normal HTTP request from the first cache. What does a False Hit look like?

The easiest situation to analyse is when two caches (say A and B) are involved neither of which uses the other as a parent. In this case, a False Hit would show up as a CACHE_DIGEST_HIT on A and NOT as a TCP_HIT on B (or vice versa). If B does not fetch the object for A then the hierarchy field will look like NONE/- (and A will have received an Access Denied or Forbidden message). This will happen if the object is not "available" on B and B does not have miss_access enabled for A (or is not acting as a parent for A). How is the cause of a False Hit determined?

Assume A requests a URL from B and receives a False Hit Using the squidclient utility PURGE the URL from A, e.g.

squidclient -m PURGE 'URL' Using the squidclient utility request the object from A, e.g.

squidclient 'URL'

The HTTP headers of the request are available. Two header types are of particular interest:

X-Cache - this shows whether an object is available or not. X-Cache-Lookup - this keeps the result of a store table lookup before refresh causing rules are checked (i.e. it indicates if the object is available before any validation would be attempted).

The X-Cache and X-Cache-Lookup headers from A should both show MISS.

If A requests the object from B (which it will if the digest lookup indicates B has it - assuming B is closest peer of course :-) then there will be another set of these headers from B.

If the X-Cache header from B shows a MISS a False Hit has occurred. This means that A thought B had an object but B tells A it does not have it available for retrieval. The reason why it is not available for retrieval is indicated by the X-Cache-Lookup header. If:

X-Cache-Lookup = MISS then either A's (version of B's) digest is out-of-date or corrupt OR a collision occurred in the digest (very small probability) OR B recently purged the object. X-Cache-Lookup = HIT then B had the object, but refresh rules (or A's max-age requirements) prevent A from getting a HIT (validation failed). Use The Source

If there is something else you need to check you can always look at the source code. The main Cache Digest functionality is organised as follows:

CacheDigest.c (debug section 70) Generic Cache Digest routines store_digest.c (debug section 71) Local Cache Digest routines peer_digest.c (debug section 72) Peer Cache Digest routines

Note that in the source the term Store Digest refers to the digest created locally. The Cache Digest code is fairly self-explanatory (once you understand how Cache Digests work): What about ICP?

COMING SOON! Is there a Cache Digest Specification?

There is now, thanks to and .

Cache Digests, as implemented in Squid 2.1.PATCH2, are described in . You'll notice the format is similar to an Internet Draft. We decided not to submit this document as a draft because Cache Digests will likely undergo some important changes before we want to try to make it a standard. Would it be possible to stagger the timings when cache_digests are retrieved from peers?

Note: The information here is current for version 2.2.

Squid already has code to spread the digest updates. The algorithm is currently controlled by a few hard-coded constants in Note that whatever you do, you still need to give Squid enough time and bandwidth to fetch all the digests. Depending on your environment, that bandwidth may be more or less than an ICP would require. Upcoming digest deltas (x10 smaller than the digests themselves) may be the only way to solve the ``big scale'' problem. Interception Caching/Proxying

How can I make my users' browsers use my cache without configuring the browsers for proxying? First, it is Getting interception caching to work requires four distinct steps: http_port 8080 httpd_accel_host virtual httpd_accel_port 80 httpd_accel_with_proxy on httpd_accel_uses_host_header on The In the The You Interception caching for Solaris, SunOS, and BSD systems

NOTE: You don't need to use IP Filter on FreeBSD. Use the built-in Install IP Filter

First, get and install the . Configure ipnat

Put these lines in /etc/ipnat.rules: # Redirect direct web traffic to local web server. rdr de0 port 80 -> port 80 tcp # Redirect everything else to squid on port 8080 rdr de0 port 80 -> port 8080 tcp

Modify your startup scripts to enable ipnat. For example, on FreeBSD it looks something like this: /sbin/modload /lkm/if_ipl.o /sbin/ipnat -f /etc/ipnat.rules chgrp nobody /dev/ipnat chmod 644 /dev/ipnat Configure Squid Squid-2

Squid-2 (after version beta25) has IP filter support built in. Simple enable it when you run ./configure --enable-ipf-transparent Add these lines to your http_port 8080 httpd_accel_host virtual httpd_accel_port 80 httpd_accel_with_proxy on httpd_accel_uses_host_header on Note, you don't have to use port 8080, but it must match whatever you used in the /etc/ipnat.rules file. Squid-1.1

Patches for Squid-1.X are available from . Add these lines to http_port 8080 httpd_accel virtual 80 httpd_accel_with_proxy on httpd_accel_uses_host_header on

Thanks to . Interception caching with Linux 2.0 and ipfwadm



Since the browser wasn't set up to use a proxy server, it uses the FTP protocol (with destination port 21) and not the required HTTP protocol. You can't setup a redirection-rule to the proxy server since the browser is speaking the wrong protocol. A similar problem occurs with gopher. Normally all proxy requests are translated by the client into the HTTP protocol, but since the client isn't aware of the redirection, this never happens.

If you can live with the side-effects, go ahead and compile your kernel with firewalling and redirection support. Here are the important parameters from /usr/src/linux/.config: # # Code maturity level options # CONFIG_EXPERIMENTAL=y # # Networking options # CONFIG_FIREWALL=y # CONFIG_NET_ALIAS is not set CONFIG_INET=y CONFIG_IP_FORWARD=y # CONFIG_IP_MULTICAST is not set CONFIG_IP_FIREWALL=y # CONFIG_IP_FIREWALL_VERBOSE is not set CONFIG_IP_MASQUERADE=y CONFIG_IP_TRANSPARENT_PROXY=y CONFIG_IP_ALWAYS_DEFRAG=y # CONFIG_IP_ACCT is not set CONFIG_IP_ROUTER=y

You may also need to enable echo 1 > /proc/sys/net/ipv4/ip_forward

Go to the page, obtain the source distribution to /etc/rc.d/rc.inet1 (Slackware) which sets up the interfaces at boot-time. The redirection should be done before any other Input-accept rule. To really make sure it worked I disabled the forwarding (masquerading) I normally do.

/etc/rc.d/rc.firewall: #!/bin/sh # rc.firewall Linux kernel firewalling rules FW=/sbin/ipfwadm # Flush rules, for testing purposes for i in I O F # A # If we enabled accounting too do ${FW} -$i -f done # Default policies: ${FW} -I -p rej # Incoming policy: reject (quick error) ${FW} -O -p acc # Output policy: accept ${FW} -F -p den # Forwarding policy: deny # Input Rules: # Loopback-interface (local access, eg, to local nameserver): ${FW} -I -a acc -S localhost/32 -D localhost/32 # Local Ethernet-interface: # Redirect to Squid proxy server: ${FW} -I -a acc -P tcp -D default/0 80 -r 8080 # Accept packets from local network: ${FW} -I -a acc -P all -S localnet/8 -D default/0 -W eth0 # Only required for other types of traffic (FTP, Telnet): # Forward localnet with masquerading (udp and tcp, no icmp!): ${FW} -F -a m -P tcp -S localnet/8 -D default/0 ${FW} -F -a m -P udp -S localnet/8 -D default/0

Here all traffic from the local LAN with any destination gets redirected to the local port 8080. Rules can be viewed like this: IP firewall input rules, default policy: reject type prot source destination ports acc all n/a acc/r tcp * -> 80 => 8080 acc all n/a acc tcp * -> *

I did some testing on Windows 95 with both Microsoft Internet Explorer 3.01 and Netscape Communicator pre-release and it worked with both browsers with the proxy-settings disabled.

At one time ${FW} -I -a rej -P tcp -S localnet/8 -D hostname/32 80 IP firewall input rules, default policy: reject type prot source destination ports acc all n/a rej tcp * -> 80 acc/r tcp * -> 80 => 8080 acc all n/a acc tcp * -> *

If you're already running a nameserver at the firewall or proxy server (which is a good idea anyway IMHO) let the workstations use this nameserver.

Additional notes from

I'm using such a setup. The only issues so far have been that: It's fairly useless to use my service providers parent caches (cache-?.www.demon.net) because by proxying squid only sees IP addresses, not host names and demon aren't generally asked for IP addresses by other users; Linux kernel 2.0.30 is a no-no as interception proxying is broken (I use 2.0.29); Client browsers must do host name lookups themselves, as they don't know they're using a proxy; The Microsoft Network won't authorize its users through a proxy, so I have to specifically *not* redirect those packets (my company is a MSN content provider). Aside from this, I get a 30-40% hit rate on a 50MB cache for 30-40 users and am quite pleased with the results.

See also . Interception caching with Linux 2.2 and ipchains


You need to configure your kernel for ipchains. Configuring Linux kernels is beyond the scope of this FAQ. One way to do it is: # cd /usr/src/linux # make menuconfig

The following shows important kernel features to include: [*] Network firewalls [ ] Socket Filtering [*] Unix domain sockets [*] TCP/IP networking [ ] IP: multicasting [ ] IP: advanced router [ ] IP: kernel level autoconfiguration [*] IP: firewalling [ ] IP: firewall packet netlink device [*] IP: always defragment (required for masquerading) [*] IP: transparent proxy support

You must include the IP: always defragment, otherwise it prevents you from using the REDIRECT chain.

You can use this script as a template for your own #!/bin/sh # rc.firewall Linux kernel firewalling rules # Leon Brooks (leon at brooks dot fdns dot net) FW=/sbin/ipchains ADD="$FW -A" # Flush rules, for testing purposes for i in I O F # A # If we enabled accounting too do ${FW} -F $i done # Default policies: ${FW} -P input REJECT # Incoming policy: reject (quick error) ${FW} -P output ACCEPT # Output policy: accept ${FW} -P forward DENY # Forwarding policy: deny # Input Rules: # Loopback-interface (local access, eg, to local nameserver): ${ADD} input -j ACCEPT -s localhost/32 -d localhost/32 # Local Ethernet-interface: # Redirect to Squid proxy server: ${ADD} input -p tcp -d 0/0 80 -j REDIRECT 8080 # Accept packets from local network: ${ADD} input -j ACCEPT -s localnet/8 -d 0/0 -i eth0 # Only required for other types of traffic (FTP, Telnet): # Forward localnet with masquerading (udp and tcp, no icmp!): ${ADD} forward -j MASQ -p tcp -s localnet/8 -d 0/0 ${ADD} forward -j MASQ -P udp -s localnet/8 -d 0/0

Also, notes that with 2.0.x kernels you don't need to enable packet forwarding, but with the 2.1.x and 2.2.x kernels using ipchains you do. Packet forwarding is enabled with the following command: echo 1 > /proc/sys/net/ipv4/ip_forward Interception caching with Linux 2.4 and netfilter

NOTE: this information comes from Daniel Kiracofe's .

To support netfilter transparent interception on Linux 2.4 Squid must be compiled with the --enable-linux-netfilter option.

To enable netwfilter support you may need to build a new kernel. Be sure to enable all of these options: Networking support Sysctl support Network packet filtering TCP/IP networking Connection tracking (Under ``IP: Netfilter Configuration'' in menuconfig) IP tables support Full NAT REDIRECT target support /proc filesystem support

You must say NO to ``Fast switching''

After building the kernel, install it and reboot.

You may need to enable packet forwarding (e.g. in your startup scripts): echo 1 > /proc/sys/net/ipv4/ip_forward

Use the iptables -t nat -A PREROUTING -i eth0 -p tcp --dport 80 -j REDIRECT --to-port 3128 Interception caching with Cisco routers


This works with at least IOS 11.1 and later I guess. Possibly earlier, as I'm no CISCO expert I can't say for sure. If your router is doing anything more complicated that shuffling packets between an ethernet interface and either a serial port or BRI port, then you should work through if this will work for you.

First define a route map with a name of proxy-redirect (name doesn't matter) and specify the next hop to be the machine Squid runs on. ! route-map proxy-redirect permit 10 match ip address 110 set ip next-hop ! Define an access list to trap HTTP requests. The second line allows the Squid host direct access so an routing loop is not formed. By carefully writing your access list as show below, common cases are found quickly and this can greatly reduce the load on your router's processor. ! access-list 110 deny tcp any any neq www access-list 110 deny tcp host any access-list 110 permit tcp any any ! Apply the route map to the ethernet interface. ! interface Ethernet0 ip policy route-map proxy-redirect ! possible bugs

notes that there is a Cisco bug relating to interception proxying using IP policy route maps, that causes NFS and other applications to break. Apparently there are two bug reports raised in Cisco, but they are not available for public dissemination.

The problem occurs with o/s packets with more than 1472 data bytes. If you try to ping a host with more than 1472 data bytes across a Cisco interface with the access-lists and ip policy route map, the icmp request will fail. The packet will be fragmented, and the first fragment is checked against the access-list and rejected - it goes the "normal path" as it is an icmp packet - however when the second fragment is checked against the access-list it is accepted (it isn't regarded as an icmp packet), and goes to the action determined by the policy route map!

notes that you may be able to get around this bug by carefully writing your access lists. If the last/default rule is to permit then this bug would be a problem, but if the last/default rule was to deny then it won't be a problem. I guess fragments, other than the first, don't have the information available to properly policy route them. Normally TCP packets should not be fragmented, at least my network runs an MTU of 1500 everywhere to avoid fragmentation. So this would affect UDP and ICMP traffic only.

Basically, you will have to pick between living with the bug or better performance. This set has better performance, but suffers from the bug: access-list 110 deny tcp any any neq www access-list 110 deny tcp host any access-list 110 permit tcp any any Conversely, this set has worse performance, but works for all protocols: access-list 110 deny tcp host any access-list 110 permit tcp any any eq www access-list 110 deny tcp any any Interception caching with LINUX 2.0.29 and CISCO IOS 11.1

Just for kicks, here's an email message posted to squid-users on how to make interception proxying work with a Cisco router and Squid running on Linux.


Here is how I have Interception proxying working for me, in an environment where my router is a Cisco 2501 running IOS 11.1, and Squid machine is running Linux 2.0.33.

Many thanks to the following individuals and the squid-users list for helping me get redirection and interception proxying working on my Cisco/Linux box. Lincoln Dale Riccardo Vratogna Mark White Henrik Nordstrom

First, here is what I added to my Cisco, which is running IOS 11.1. In IOS 11.1 the route-map command is "process switched" as opposed to the faster "fast-switched" route-map which is found in IOS 11.2 and later. You may wish to be running IOS 11.2. I am running 11.1, and have had no problems with my current load of about 150 simultaneous connections to squid.: ! interface Ethernet0 description To Office Ethernet ip address no ip directed-broadcast no ip mroute-cache ip policy route-map proxy-redir ! access-list 110 deny tcp host any eq www access-list 110 permit tcp any any eq www route-map proxy-redir permit 10 match ip address 110 set ip next-hop

So basically from above you can see I added the "route-map" declaration, and an access-list, and then turned the route-map on under int e0 "ip policy route-map proxy-redir"

ok, so the Cisco is taken care of at this point. The host above:, is the ip number of my squid host.

My squid box runs Linux, so I had to do the following on it:


You will need Firewalling and Transparent Proxy turned on at a minimum.

Then some ipfwadm stuff: # Accept all on loopback ipfwadm -I -a accept -W lo # Accept my own IP, to prevent loops (repeat for each interface/alias) ipfwadm -I -a accept -P tcp -D 80 # Send all traffic destined to port 80 to Squid on port 3128 ipfwadm -I -a accept -P tcp -D 0/0 80 -r 3128

it accepts packets on port 80 (redirected from the Cisco), and redirects them to 3128 which is the port my squid process is sitting on. I put all this in /etc/rc.d/rc.local

I am using with installed. You will want to install this patch if using a setup similar to mine. The cache is trying to connect to itself...


I think almost everyone who have tried to build a interception proxy setup have been bitten by this one.

Measures you can take: Deny Squid from fetching objects from itself (using ACL lists). Apply a small patch that prevents Squid from looping infinitely (available from ) Don't run Squid on port 80, and redirect port 80 not destined for the local machine to Squid (redirection == ipfilter/ipfw/ipfadm). This avoids the most common loops. If you are using ipfilter then you should also use transproxyd in front of Squid. Squid does not yet know how to interface to ipfilter (patches are welcome: squid-bugs@squid-cache.org). Interception caching with FreeBSD

by Duane Wessels

I set out yesterday to make interception caching work with Squid and FreeBSD. It was, uh, fun.

It was relatively easy to configure a cisco to divert port 80 packets to my FreeBSD box. Configuration goes something like this: access-list 110 deny tcp host any eq www access-list 110 permit tcp any any eq www route-map proxy-redirect permit 10 match ip address 110 set ip next-hop int eth2/0 ip policy route-map proxy-redirect Here, is the IP address of the FreeBSD cache machine.

Once I have packets going to the FreeBSD box, I need to get the kernel to deliver them to Squid. I started on FreeBSD-2.2.7, and then downloaded . This was a dead end for me. The IPFilter distribution includes patches to the FreeBSD kernel sources, but many of these had conflicts. Then I noticed that the IPFilter page says ``It comes as a part of [FreeBSD-2.2 and later].'' Fair enough. Unfortunately, you can't hijack connections with the FreeBSD-2.2.X IPFIREWALL code ( FreeBSD-3.0 has much better support for connection hijacking, so I suggest you start with that. You need to build a kernel with the following options: options IPFIREWALL options IPFIREWALL_FORWARD

Next, its time to configure the IP firewall rules with /etc/rc.local just to be able to use the machine on my network: ipfw add 60000 allow all from any to any But we're still not hijacking connections. To accomplish that, add these rules: ipfw add 49 allow tcp from to any ipfw add 50 fwd tcp from any to any 80 The second line (rule 50) is the one which hijacks the connection. The first line makes sure we never hit rule 50 for traffic originated by the local machine. This prevents forwarding loops.

Note that I am not changing the port number here. That is, port 80 packets are simply diverted to Squid on port 80. My Squid configuration is: http_port 80 httpd_accel_host virtual httpd_accel_port 80 httpd_accel_with_proxy on httpd_accel_uses_host_header on

If you don't want Squid to listen on port 80 (because that requires root privileges) then you can use another port. In that case your ipfw redirect rule looks like: ipfw add 50 fwd,3128 tcp from any to any 80 and the http_port 3128 httpd_accel_host virtual httpd_accel_port 80 httpd_accel_with_proxy on httpd_accel_uses_host_header on Interception caching with ACC Tigris digital access server


This is to do with configuring interception proxy for an ACC Tigris digital access server (like a CISCO 5200/5300 or an Ascend MAX 4000). I've found that doing this in the NAS reduces traffic on the LAN and reduces processing load on the CISCO. The Tigris has ample CPU for filtering.

Step 1 is to create filters that allow local traffic to pass. Add as many as needed for all of your address ranges. ADD PROFILE IP FILTER ENTRY local1 INPUT NORMAL ADD PROFILE IP FILTER ENTRY local2 INPUT NORMAL

Step 2 is to create a filter to trap port 80 traffic. ADD PROFILE IP FILTER ENTRY http INPUT = 0x6 D= 80 NORMAL

Step 3 is to set the "APPLICATION_ID" on port 80 traffic to 80. This causes all packets matching this filter to have ID 80 instead of the default ID of 0. SET PROFILE IP FILTER APPLICATION_ID http 80

Step 4 is to create a special route that is used for packets with "APPLICATION_ID" set to 80. The routing engine uses the ID to select which routes to use. ADD IP ROUTE ENTRY PROXY-IP 1 SET IP ROUTE APPLICATION_ID PROXY-IP 80

Step 5 is to bind everything to a filter ID called transproxy. List all local filters first and the http one last. ADD PROFILE ENTRY transproxy local1 local2 http

With this in place use your RADIUS server to send back the ``Framed-Filter-Id = transproxy'' key/value pair to the NAS.

You can check if the filter is being assigned to logins with the following command: display profile port table ``Connection reset by peer'' and Cisco policy routing

has tracked down the cause of unusual ``connection reset by peer'' messages when using Cisco policy routing to hijack HTTP requests.

When the network link between router and the cache goes down for just a moment, the packets that are supposed to be redirected are instead sent out the default route. If this happens, a TCP ACK from the client host may be sent to the origin server, instead of being diverted to the cache. The origin server, upon receiving an unexpected ACK packet, sends a TCP RESET back to the client, which aborts the client's request.

To work around this problem, you can install a static route to the ip route Null0 250 This appears to cause the correct behaviour. WCCP - Web Cache Coordination Protocol

Contributors: , and . Does Squid support WCCP?

CISCO's Web Cache Coordination Protocol V1.0 is supported in squid 2.3 and later. support WCCP V2.0. Now that WCCP V2 is an open protocol, Squid may be able to support it in the future. Configuring your Router

There are two different methods of configuring WCCP on CISCO routers. The first method is for routers that only support V1.0 of the protocol. The second is for routers that support both. IOS Version 11.x

It is possible that later versions of IOS 11.x will support V2.0 of the protocol. If that is the case follow the 12.x instructions. Several people have reported that the squid implimentation of WCCP does not work with their 11.x routers. If you experience this please mail the debug output from your router to conf t wccp enable ! interface [Interface carrying Outgoing Traffic]x/x ! ip wccp web-cache redirect ! CTRL Z write mem IOS Version 12.x

Some of the early versions of 12.x do not have the 'ip wccp version' command. You will need to upgrade your IOS version to use V1.0.

You will need to be running at least IOS Software Release conf t ip wccp version 1 ip wccp web-cache redirect-list 150 ! interface [Interface carrying Outgoing/Incoming Traffic]x/x ip wccp web-cache redirect out|in ! CTRL Z write mem

Replace 150 with an access list number (either standard or extended) which lists IP addresses which you do not wish to be transparently redirected to your cache. Otherwise simply user the word 'redirect' on it's own to redirect traffic from all sources to all destinations. IOS 12.x problems

Some people report problems with WCCP and IOS 12.x. They see truncated or fragmented GRE packets arriving at the cache. Apparently it works if you disable Cisco Express Forwarding for the interface: conf t ip cef # some systems may already have 'ip cef global' int Ethernet 0/0 (or int FastEthernet 0/0 or other internal interface) no ip route-cache cef CTRL Z

This may well be fixed in later releases of IOS. Configuring FreeBSD

FreeBSD first needs to be configured to receive and strip the GRE encapsulation from the packets from the router. To do this you will need to patch and recompile your kernel. The steps depend on your kernel version. FreeBSD-3.x

Apply the : # cd /usr/src # patch -s < /tmp/gre.patch Download . Save this file as /usr/src/sys/netinet/gre.c. Add "options GRE" to your kernel config file and rebuild your kernel. Note, the . FreeBSD-4.0 through 4.7

The procedure is nearly identical to the above for 3.x, but the source files are a little different. Apply the most appropriate patch file from the list of . Download . Save this file as /usr/src/sys/netinet/gre.c. Add "options GRE" to your kernel config file and rebuild your kernel. Note, the . FreeBSD-4.8 and later

The operating system now comes standard with some GRE support. You need to make a kernel with the GRE code enabled: pseudo-device gre

And then configure the tunnel so that the router's GRE packets are accepted: # ifconfig gre0 create # ifconfig gre0 $squid_ip $router_ip netmask up # ifconfig gre0 tunnel $squid_ip $router_ip # route delete $router_ip Configuring Linux 2.2

Al Blake has written a .

There are currently two methods for supporting WCCP with Linux 2.2. A specific purpose module. Or the standard Linux GRE tunneling driver. People have reported difficulty with the standard GRE tunneling driver, however it does allow GRE functionality other than WCCP. You should choose the method that suits your enviroment. Standard Linux GRE Tunnel

Linux 2.2 kernels already support GRE, as long as the GRE module is compiled into the kernel.

Ensure that the GRE code is either built as static or as a module by chosing the appropriate option in your kernel config. Then rebuild your kernel. If it is a module you will need to: modprobe ip_gre The next step is to tell Linux to establish an IP tunnel between the router and your host. Daniele Orlandi reports that you have to give the gre1 interface an address, but any old address seems to work. iptunnel add gre1 mode gre remote <Router-IP> local <Host-IP> dev <interface> ifconfig gre1 up <Router-IP> is the IP address of your router that is intercepting the HTTP packets. <Host-IP> is the IP address of your cache, and <interface> is the network interface that receives those packets (probably eth0). Joe Cooper's Patch

Joe Cooper has a patch for Linux 2.2.18 kernel on his . WCCP Specific Module

This module is not part of the standard Linux distributon. It needs to be compiled as a module and loaded on your system to function. Do not attempt to build this in as a static part of your kernel.

Download the and compile it as you would any Linux network module.

Copy the module to /lib/modules/kernel-version/ipv4/ip_wccp.o. Edit /lib/modules/kernel-version/modules.dep and add: /lib/modules/kernel-version/ipv4/ip_wccp.o:

Finally you will need to load the module: modprobe ip_wccp Common Steps

The machine should now be striping the GRE encapsulation from any packets recieved and requeuing them. The system will also need to be configured for interception proxying, either with or with . Configuring Others

If you have managed to configuring your operating system to support WCCP with Squid please contact us with the details so we may share them with others. Can someone tell me what version of cisco IOS WCCP is added in?

IOS releases: 11.1(19?)CA/CC or later 11.2(14)P or later 12.0(anything) or later What about WCCPv2?

Cisco has published WCCPv2 as an (expired Jan 2001). There is a ongoing project at the website aiming to add support for WCCPv2 and at the time of writing this patch provides at least the same functionality as WCCPv1. Interception caching with Foundry L4 switches

by .

First, configure Squid for interception caching as detailed at the .

Next, configure the Foundry layer 4 switch to redirect traffic to your Squid box or boxes. By default, the Foundry redirects to port 80 of your squid box. This can be changed to a different port if needed, but won't be covered here.

In addition, the switch does a "health check" of the port to make sure your squid is answering. If you squid does not answer, the switch defaults to sending traffic directly thru instead of redirecting it. When the Squid comes back up, it begins redirecting once again.

This example assumes you have two squid caches: squid1.foo.com squid2.foo.com

We will assume you have various workstations, customers, etc, plugged into the switch for which you want them to be intercepted and sent to Squid. The squid caches themselves should be plugged into the switch as well. Only the interface that the router is connected to is important. Where you put the squid caches or other connections does not matter.

This example assumes your router is plugged into interface Enter configuration mode: telnet@ServerIron#conf t Configure each squid on the Foundry: telnet@ServerIron(config)# server cache-name squid1 telnet@ServerIron(config)# server cache-name squid2 Add the squids to a cache-group: telnet@ServerIron(config)#server cache-group 1 telnet@ServerIron(config-tc-1)#cache-name squid1 telnet@ServerIron(config-tc-1)#cache-name squid2 Create a policy for caching http on a local port telnet@ServerIron(config)# ip policy 1 cache tcp http local Enable that policy on the port connected to your router telnet@ServerIron(config)#int e 17 telnet@ServerIron(config-if-17)# ip-policy 1

Since all outbound traffic to the Internet goes out interface The default port to redirect to can be changed. The load balancing algorithm used can be changed (Least Used, Round Robin, etc). Ports can be exempted from caching if needed. Access Lists can be applied so that only certain source IP Addresses are redirected, etc. This information was left out of this document since this was just a quick howto that would apply for most people, not meant to be a comprehensive manual of how to configure a Foundry switch. I can however revise this with any information necessary if people feel it should be included. Can I use No, you cannot. With interception proxying, the client thinks it is talking to an origin server and would never send the SNMP

Contributors: . Does Squid support SNMP?

True SNMP support is available in squid 2 and above. A significant change in the implimentation occured starting with the development 2.2 code. Therefore there are two sets of instructions on how to configure SNMP in squid, please make sure that you follow the correct one. Enabling SNMP in Squid

To use SNMP, it must first be enabled with the ./configure --enable-snmp [ ... other configure options ] Next, recompile after cleaning the source tree : make clean make all make install Once the compile is completed and the new binary is installed the Configuring Squid 2.2

To configure SNMP first specify a list of communities that you would like to allow access by using a standard acl of the form: acl aclname snmp_community string For example: acl snmppublic snmp_community public acl snmpjoebloggs snmp_community joebloggs This creates two acl's, with two different communities, public and joebloggs. You can name the acl's and the community strings anything that you like.

To specify the port that the agent will listen on modify the "snmp_port" parameter, it is defaulted to 3401. The port that the agent will forward requests that can not be furfilled by this agent to is set by "forward_snmpd_port" it is defaulted to off. It must be configured for this to work. Remember that as the requests will be originating from this agent you will need to make sure that you configure your access accordingly.

To allow access to Squid's SNMP agent, define an snmp_access allow snmppublic localhost snmp_access deny all The above will allow anyone on the localhost who uses the community If you do not define any Finally squid allows to you to configure the address that the agent will bind to for incomming and outgoing traffic. These are defaulted to, changing these will cause the agent to bind to a specific address on the host, rather than the default which is all. snmp_incoming_address snmp_outgoing_address Configuring Squid 2.1

Prior to Squid 2.1 the SNMP code had a number of issues with the ACL's. If you are a frequent user of SNMP with Squid, please upgrade to 2.2 or higher.

A sort of default, working configuration is: snmp_port 3401 snmp_mib_path /local/squid/etc/mib.txt snmp_agent_conf view all .1.3.6 included snmp_agent_conf view squid .1.3.6 included snmp_agent_conf user squid - all all public snmp_agent_conf user all all all all squid snmp_agent_conf community public squid squid snmp_agent_conf community readwrite all all

Note that for security you are advised to restrict SNMP access to your caches. You can do this easily as follows: acl snmpmanagementhosts snmp_acl public deny all !snmpmanagementhosts snmp_acl readwrite deny all You must follow these instructions for 2.1 and below exactly or you are likely to have problems. The parser has some issues which have been corrected in 2.2. How can I query the Squid SNMP Agent

You can test if your Squid supports SNMP with the ). Note that you have to specify the SNMP port, which in Squid defaults to 3401. snmpwalk -p 3401 hostname communitystring . If it gives output like: enterprises.nlanr.squid.cacheSystem.cacheSysVMsize = 7970816 enterprises.nlanr.squid.cacheSystem.cacheSysStorage = 2796142 enterprises.nlanr.squid.cacheSystem.cacheUptime = Timeticks: (766299) 2:07:42.99 then it is working ok, and you should be able to make nice statistics out of it.

For an explanation of what every string (OID) does, you should refer to the . What can I use SNMP and Squid for?

There are a lot of things you can do with SNMP and Squid. It can be useful in some extent for a longer term overview of how your proxy is doing. It can also be used as a problem solver. For example: how is it going with your filedescriptor usage? or how much does your LRU vary along a day. Things you can't monitor very well normally, aside from clicking at the cachemgr frequently. Why not let MRTG do it for you? How can I use SNMP with Squid?

There are a number of tools that you can use to monitor Squid via SNMP. Many people use MRTG. Another good combination is plus . You might be able to find more information at the or Where can I get more information/discussion about Squid and SNMP?

General Discussion: These messages are .

Subscriptions should be sent to: . Monitoring Squid with MRTG

Some people use to query Squid through its SNMP interface.

To get instruction on using MRTG with Squid please visit these pages: by DFN-Cache by Guido Serassio by Visolve Desire II caching workshop session by Matija Grabnar by The National Janet Web Cache Service

Further examples of Squid MRTG configurations can be found here: from MRTG from MRTG by Glenn Chisholm by Joakim Recht Squid version 2 What are the new features?

persistent connections. Lower VM usage; in-transit objects are not held fully in memory. Totally independent swap directories. Customizable error texts. FTP supported internally; no more ftpget. Asynchronous disk operations (optional, requires pthreads library). Internal icons for FTP and gopher directories. snprintf() used everywhere instead of sprintf(). SNMP. Routing requests based on AS numbers. ...and many more! How do I configure 'ssl_proxy' now?

By default, Squid connects directly to origin servers for SSL requests. But if you must force SSL requests through a parent, first tell Squid it can not go direct for SSL: acl SSL method CONNECT never_direct allow SSL With this in place, Squid cache_peer parent1 parent 3128 3130 cache_peer parent2 parent 3128 3130 cache_peer_access parent2 allow !SSL The above lines tell Squid to NOT use Logfile rotation doesn't work with Async I/O

It is a know limitation when using Async I/O on Linux. The Linux Threads package steals (uses internally) the SIGUSR1 signal that squid uses to rotate logs.

In order to not disturb the threads package SIGUSR1 use is disabled in Squid when threads is enabled on Linux. Adding a new cache disk

Simply add your new Squid 2 performs badly on Linux


You may have enabled Asyncronous I/O with the You should also know that How do I configure proxy authentication with Squid-2?

For Squid-2, the implementation and configuration has changed. Authentication is now handled via external processes. Arjan's describes how to set it up. Some simple instructions are given below as well. We assume you have configured an ACL entry with proxy_auth, for example: acl foo proxy_auth REQUIRED http_access allow foo You will need to compile and install an external authenticator program. Most people will want to use auth_modules/NCSA directory. % cd auth_modules/NCSA % make % make install You should now have an You may need to create a password file. If you have been using proxy authentication before, you probably already have such a file. You can get from our server. Pick a pathname for your password file. We will assume you will want to put it in the same directory as your squid.conf. Configure the external authenticator in auth_param basic /usr/local/squid/bin/ncsa_auth /usr/local/squid/etc/passwd

After all that, you should be able to start up Squid. If we left something out, or haven't been clear enough, please let us know (squid-faq@squid-cache.org). Why does proxy-auth reject all users after upgrading from Squid-2.1 or earlier?

The ACL for proxy-authentication has changed from: acl foo proxy_auth timeout to: acl foo proxy_auth username Please update your ACL appropriately - a username of auth_param basic credentialsttl timeout Delay Pools

by .

The information here is current for version 2.2. It is strongly recommended that you use at least Squid 2.2 if you wish to use delay pools.

Delay pools provide a way to limit the bandwidth of certain requests based on any list of criteria. The idea came from a Western Australian university who wanted to restrict student traffic costs (without affecting staff traffic, and still getting cache and local peering hits at full speed). There was some early Squid 1.0 code by Central Network Services at Murdoch University, which I then developed (at the University of Western Australia) into a much more complex patch for Squid 1.0 called ``DELAY_HACK.'' I then tried to code it in a much cleaner style and with slightly more generic options than I personally needed, and called this ``delay pools'' in Squid 2. I almost completely recoded this in Squid 2.2 to provide the greater flexibility requested by people using the feature.

To enable delay pools features in Squid 2.2, you must use the --enable-delay-pools configure option before compilation.

Terminology for this FAQ entry:

Delay pools allows you to limit traffic for clients or client groups, with various features: can specify peer hosts which aren't affected by delay pools, ie, local peering or other 'free' traffic (with the no-delay peer option). delay behavior is selected by ACLs (low and high priority traffic, staff vs students or student vs authenticated student or so on). each group of users has a number of buckets, a bucket has an amount coming into it in a second and a maximum amount it can grow to; when it reaches zero, objects reads are deferred until one of the object's clients has some traffic allowance. any number of pools can be configured with a given class and any set of limits within the pools can be disabled, for example you might only want to use the aggregate and per-host bucket groups of class 3, not the per-network one.

This allows options such as creating a number of class 1 delay pools and allowing a certain amount of bandwidth to given object types (by using URL regular expressions or similar), and many other uses I'm sure I haven't even though of beyond the original fair balancing of a relatively small traffic allocation across a large number of users.

There are some limitations of delay pools: delay pools are incompatible with slow aborts; quick abort should be set fairly low to prevent objects being retrived at full speed once there are no clients requesting them (as the traffic allocation is based on the current clients, and when there are no clients attached to the object there is no way to determine the traffic allocation). delay pools only limits the actual data transferred and is not inclusive of overheads such as TCP overheads, ICP, DNS, icmp pings, etc. it is possible for one connection or a small number of connections to take all the bandwidth from a given bucket and the other connections to be starved completely, which can be a major problem if there are a number of large objects being transferred and the parameters are set in a way that a few large objects will cause all clients to be starved (potentially fixed by a currently experimental patch). How can I limit Squid's total bandwidth to, say, 512 Kbps?

acl all src # might already be defined delay_pools 1 delay_class 1 1 delay_access 1 allow all delay_parameters 1 64000/64000 # 512 kbits == 64 kbytes per second For an explanation of these tags please see the configuration file.

The 1 second buffer (max = restore = 64kbytes/sec) is because a limit is requested, and no responsiveness to a busrt is requested. If you want it to be able to respond to a burst, increase the aggregate_max to a larger value, and traffic bursts will be handled. It is recommended that the maximum is at least twice the restore value - if there is only a single object being downloaded, sometimes the download rate will fall below the requested throughput as the bucket is not empty when it comes to be replenished. How to limit a single connection to 128 Kbps?

You can not limit a single HTTP request's connection speed. You can limit individual hosts to some bandwidth rate. To limit a specific host, define an acl for that host and use the example above. To limit a group of hosts, then you must use a delay pool of class 2 or 3. For example: acl only128kusers src acl all src delay_pools 1 delay_class 1 3 delay_access 1 allow only128kusers delay_access 1 deny all delay_parameters 1 64000/64000 -1/-1 16000/64000 For an explanation of these tags please see the configuration file. The above gives a solution where a cache is given a total of 512kbits to operate in, and each IP address gets only 128kbits out of that pool. How do you personally use delay pools?

We have six local cache peers, all with the options 'proxy-only no-delay' since they are fast machines connected via a fast ethernet and microwave (ATM) network.

For our local access we use a dstdomain ACL, and for delay pool exceptions we use a dst ACL as well since the delay pool ACL processing is done using "fast lookups", which means (among other things) it won't wait for a DNS lookup if it would need one.

Our proxy has two virtual interfaces, one which requires student authentication to connect from machines where a department is not paying for traffic, and one which uses delay pools. Also, users of the main Unix system are allowed to choose slow or fast traffic, but must pay for any traffic they do using the fast cache. Ident lookups are disabled for accesses through the slow cache since they aren't needed. Slow accesses are delayed using a class 3 delay pool to give fairness between departments as well as between users. We recognize users of Lynx on the main host are grouped together in one delay bucket but they are mostly viewing text pages anyway, so this isn't considered a serious problem. If it was we could take those hosts into a class 1 delay pool and give it a larger allocation.

I prefer using a slow restore rate and a large maximum rate to give preference to people who are looking at web pages as their individual bucket fills while they are reading, and those downloading large objects are disadvantaged. This depends on which clients you believe are more important. Also, one individual 8 bit network (a residential college) have paid extra to get more bandwidth.

The relevant parts of my configuration file are (IP addresses, etc, all changed): # ACL definitions # Local network definitions, domains a.net, b.net acl LOCAL-NET dstdomain a.net b.net # Local network; nets 64 - 127. Also nearby network class A, 10. acl LOCAL-IP dst # Virtual i/f used for slow access acl virtual_slowcache myip # All permitted slow access, nets 96 - 127 acl slownets src # Special 'fast' slow access, net 123 acl fast_slow src # User hosts acl my_user_hosts src # "All" ACL acl all src # Don't need ident lookups for billing on (free) slow cache ident_lookup_access allow my_user_hosts !virtual_slowcache ident_lookup_access deny all # Security access checks http_access [...] # These people get in for slow cache access http_access allow virtual_slowcache slownets http_access deny virtual_slowcache # Access checks for main cache http_access [...] # Delay definitions (read config file for clarification) delay_pools 2 delay_initial_bucket_level 50 delay_class 1 3 delay_access 1 allow virtual_slowcache !LOCAL-NET !LOCAL-IP !fast_slow delay_access 1 deny all delay_parameters 1 8192/131072 1024/65536 256/32768 delay_class 2 2 delay_access 2 allow virtual_slowcache !LOCAL-NET !LOCAL-IP fast_slow delay_access 2 deny all delay_parameters 2 2048/65536 512/32768

The same code is also used by a some of departments using class 2 delay pools to give them more flexibility in giving different performance to different labs or students. Where else can I find out about delay pools?

This is also pretty well documented in the configuration file, with examples. Since people seem to loose their config files, here's a copy of the relevant section. # DELAY POOL PARAMETERS (all require DELAY_POOLS compilation option) # ----------------------------------------------------------------------------- # TAG: delay_pools # This represents the number of delay pools to be used. For example, # if you have one class 2 delay pool and one class 3 delays pool, you # have a total of 2 delay pools. # # To enable this option, you must use --enable-delay-pools with the # configure script. #delay_pools 0 # TAG: delay_class # This defines the class of each delay pool. There must be exactly one # delay_class line for each delay pool. For example, to define two # delay pools, one of class 2 and one of class 3, the settings above # and here would be: # #delay_pools 2 # 2 delay pools #delay_class 1 2 # pool 1 is a class 2 pool #delay_class 2 3 # pool 2 is a class 3 pool # # The delay pool classes are: # # class 1 Everything is limited by a single aggregate # bucket. # # class 2 Everything is limited by a single aggregate # bucket as well as an "individual" bucket chosen # from bits 25 through 32 of the IP address. # # class 3 Everything is limited by a single aggregate # bucket as well as a "network" bucket chosen # from bits 17 through 24 of the IP address and a # "individual" bucket chosen from bits 17 through # 32 of the IP address. # # NOTE: If an IP address is a.b.c.d # -> bits 25 through 32 are "d" # -> bits 17 through 24 are "c" # -> bits 17 through 32 are "c * 256 + d" # TAG: delay_access # This is used to determine which delay pool a request falls into. # The first matched delay pool is always used, ie, if a request falls # into delay pool number one, no more delay are checked, otherwise the # rest are checked in order of their delay pool number until they have # all been checked. For example, if you want some_big_clients in delay # pool 1 and lotsa_little_clients in delay pool 2: # #delay_access 1 allow some_big_clients #delay_access 1 deny all #delay_access 2 allow lotsa_little_clients #delay_access 2 deny all # TAG: delay_parameters # This defines the parameters for a delay pool. Each delay pool has # a number of "buckets" associated with it, as explained in the # description of delay_class. For a class 1 delay pool, the syntax is: # #delay_parameters pool aggregate # # For a class 2 delay pool: # #delay_parameters pool aggregate individual # # For a class 3 delay pool: # #delay_parameters pool aggregate network individual # # The variables here are: # # pool a pool number - ie, a number between 1 and the # number specified in delay_pools as used in # delay_class lines. # # aggregate the "delay parameters" for the aggregate bucket # (class 1, 2, 3). # # individual the "delay parameters" for the individual # buckets (class 2, 3). # # network the "delay parameters" for the network buckets # (class 3). # # A pair of delay parameters is written restore/maximum, where restore is # the number of bytes (not bits - modem and network speeds are usually # quoted in bits) per second placed into the bucket, and maximum is the # maximum number of bytes which can be in the bucket at any time. # # For example, if delay pool number 1 is a class 2 delay pool as in the # above example, and is being used to strictly limit each host to 64kbps # (plus overheads), with no overall limit, the line is: # #delay_parameters 1 -1/-1 8000/8000 # # Note that the figure -1 is used to represent "unlimited". # # And, if delay pool number 2 is a class 3 delay pool as in the above # example, and you want to limit it to a total of 256kbps (strict limit) # with each 8-bit network permitted 64kbps (strict limit) and each # individual host permitted 4800bps with a bucket maximum size of 64kb # to permit a decent web page to be downloaded at a decent speed # (if the network is not being limited due to overuse) but slow down # large downloads more significantly: # #delay_parameters 2 32000/32000 8000/8000 600/8000 # # There must be one delay_parameters line for each delay pool. # TAG: delay_initial_bucket_level (percent, 0-100) # The initial bucket percentage is used to determine how much is put # in each bucket when squid starts, is reconfigured, or first notices # a host accessing it (in class 2 and class 3, individual hosts and # networks only have buckets associated with them once they have been # "seen" by squid). # #delay_initial_bucket_level 50 Can I preserve my cache when upgrading from 1.1 to 2?

At the moment we do not have a script which will convert your cache contents from the 1.1 to the Squid-2 format. If enough people ask for one, then somebody will probably write such a script.

If you like, you can configure a new Squid-2 cache with your old Squid-1.1 cache as a sibling. After a few days, weeks, or however long you want to wait, shut down the old Squid cache. If you want to force-load your new cache with the objects from the old cache, you can try something like this: Install Squid-2 and configure it to have the same amount of disk space as your Squid-1 cache, even if there is not currently that much space free. Configure Squid-2 with Squid-1 as a parent cache. You might want to enable Enable the on Squid-1. Set the refresh rules on Squid-1 to be very liberal so that it does not generate IMS requests for cached objects. Create a list of all the URLs in the Squid-1 cache. These can be extracted from the access.log, store.log and swap logs. For every URL in the list, request the URL from Squid-2, and then immediately send a PURGE request to Squid-1. Eventually Squid-2 will have all the objects, and Squid-1 will be empty. Customizable Error Messages

Squid-2 lets you customize your error messages. The source distribution includes error messages in different languages. You can select the language with the configure option: --enable-err-language=lang

Furthermore, you can rewrite the error message template files if you like. This list describes the tags which Squid will insert into the messages: The Squid default signature is added automatically unless %s is used in the error page. To change the signature you must manually append the signature to each error page.

The default signature reads like: <BR clear="all"> <HR noshade size="1px"> <ADDRESS> Generated %T by %h (%s) </ADDRESS> </BODY></HTML> My squid.conf from version 1.1 doesn't work!

Yes, a number of configuration directives have been renamed. Here are some of them: acl Uncachable url_regex cgi ? no_cache deny Uncachable acl that-AS dst_as 1241 cache_peer_access thatcache.thatdomain.net allow that-AS cache_peer_access thatcache.thatdomain.net deny all This example sends requests to your peer connect_timeout 120 seconds read_timeout 15 minutes httpd-accelerator mode What is the httpd-accelerator mode?

Occasionally people have trouble understanding accelerators and proxy caches, usually resulting from mixed up interpretations of "incoming" and ``outgoing" data. I think in terms of requests (i.e., an outgoing request is from the local site out to the big bad Internet). The data received in reply is incoming, of course. Others think in the opposite sense of ``a request for incoming data".

An accelerator caches incoming requests for outgoing data (i.e., that which you publish to the world). It takes load away from your HTTP server and internal network. You move the server away from port 80 (or whatever your published port is), and substitute the accelerator, which then pulls the HTTP data from the ``real" HTTP server (only the accelerator needs to know where the real server is). The outside world sees no difference (apart from an increase in speed, with luck).

Quite apart from taking the load of a site's normal web server, accelerators can also sit outside firewalls or other network bottlenecks and talk to HTTP servers inside, reducing traffic across the bottleneck and simplifying the configuration. Two or more accelerators communicating via ICP can increase the speed and resilience of a web service to any single failure.

The Squid redirector can make one accelerator act as a single front-end for multiple servers. If you need to move parts of your filesystem from one server to another, or if separately administered HTTP servers should logically appear under a single URL hierarchy, the accelerator makes the right thing happen.

If you wish only to cache the ``rest of the world" to improve local users browsing performance, then accelerator mode is irrelevant. Sites which own and publish a URL hierarchy use an accelerator to improve other sites' access to it. Sites wishing to improve their local users' access to other sites' URLs use proxy caches. Many sites, like us, do both and hence run both.

Measurement of the Squid cache and its Harvest counterpart suggest an order of magnitude performance improvement over CERN or other widely available caching software. This order of magnitude performance improvement on hits suggests that the cache can serve as an httpd accelerator, a cache configured to act as a site's primary httpd server (on port 80), forwarding references that miss to the site's real httpd (on port 81).

In such a configuration, the web administrator renames all non-cachable URLs to the httpd's port (81). The cache serves references to cachable objects, such as HTML pages and GIFs, and the true httpd (on port 81) serves references to non-cachable objects, such as queries and cgi-bin programs. If a site's usage characteristics tend toward cachable objects, this configuration can dramatically reduce the site's web workload.

Note that it is best not to run a single How do I set it up?

First, you have to tell Squid to listen on port 80 (usually), so set the 'http_port' option: http_port 80

Next, you need to move your normal HTTP server to another port and/or another machine. If you want to run your HTTP server on the same machine, then it can not also use port 80 (except see the next FAQ entry below). A common choice is port 81. Configure squid as follows: httpd_accel_host localhost httpd_accel_port 81 Alternatively, you could move the HTTP server to another machine and leave it on port 80: httpd_accel_host otherhost.foo.com httpd_accel_port 80

You should now be able to start Squid and it will serve requests as a HTTP server.

If you are using Squid has an accelerator for a virtual host system, then you need to specify httpd_accel_host virtual

Finally, if you want Squid to also accept httpd_accel_with_proxy on When using an httpd-accelerator, the port number for redirects is wrong

Yes, this is because you probably moved your real httpd to port 81. When your httpd issues a redirect message (e.g. 302 Moved Temporarily), it knows it is not running on the standard port (80), so it inserts How can you fix this?

One way is to leave your httpd running on port 80, but bind the httpd socket to a you can do it like this in Port 80 BindAddress Then, in your httpd_accel_host httpd_accel_port 80

Note, you probably also need to add an /etc/hosts entry of for your server hostname. Otherwise, Squid may get stuck in a forwarding loop. Related Software Clients Wget

is a command-line Web client. It supports HTTP and FTP URLs, recursive retrievals, and HTTP proxies. echoping

If you want to test your Squid cache in batch (from a cron command, for instance), you can use the program, which will tell you (in plain text or via an exit code) if the cache is up or not, and will indicate the response times. Logfile Analysis

Rather than maintain the same list in two places, please see the page on the Web server. Configuration Tools 3Dhierarchy.pl

Kenichi Matsui has a simple perl script which generates a 3D hierarchy map (in VRML) from squid.conf. . Squid add-ons transproxy

is a program used in conjunction with the Linux Transparent Proxy networking feature, and ipfwadm, to intercept HTTP and other requests. Transproxy is written by . Iain's redirector package

A from to allow Intranet (restricted) or Internet (full) access with URL deny and redirection for sites that are not deemed acceptable for a userbase all via a single proxy port. Junkbusters

Corp has a copyleft privacy-enhancing, ad-blocking proxy server which you can use in conjunction with Squid. Squirm

is a configurable, efficient redirector for Squid by . Features: Very fast Virtually no memory usage It can re-read it's config files while running by sending it a HUP signal Interactive test mode for checking new configs Full regular expression matching and replacement Config files for patterns and IP addresses. If you mess up the config file, Squirm runs in Dodo Mode so your squid keeps working :-) chpasswd.cgi

has adapated the Apache's into a CGI program called . jesred

by . squidGuard

is a free (GPL), flexible and efficient filter and redirector program for squid. It lets you define multiple access rules with different restrictions for different user groups on a squid cache. squidGuard uses squid standard redirector interface. Central Squid Server

The Smart Neighbour [URL disappeared] (or 'Central Squid Server' - CSS) is a cut-down version of Squid without HTTP or object caching functionality. The CSS deals only with ICP messages. Instead of caching objects, the CSS records the availability of objects in each of its neighbour caches. Caches that have smart neighbours update each smart neighbour with the status of their cache by sending ICP_STORE_NOTIFY/ICP_RELEASE_NOTIFY messages upon storing/releasing an object from their cache. The CSS maintains an up to date 'object map' recording the availability of objects in its neighbouring caches. Cerberian content filter (subscription service)

The is a very flexible URL rating system with full Squid integration provided by . The service requires a license (priced by the number of seats) but evaluation licenses are available. Ident Servers

For , , and . DISKD What is DISKD?

DISKD refers to some features in Squid-2.4 and later to improve Disk I/O performance. The basic idea is that each Does it perform better?

Yes. We benchmarked Squid-2.4 with DISKD at the . The results are also described . At the bakeoff, we got 160 req/sec with diskd. Without diskd, we'd have gotten about 40 req/sec. How do I use it?

You need to run Squid version or later. Your operating system must support message queues, and shared memory.

To configure Squid for DISKD, use the % ./configure --enable-storeio=diskd,ufs FATAL: Unknown cache_dir type 'diskd'

You didn't put If I use DISKD, do I have to wipe out my current cache?

No. Diskd uses the same storage scheme as the standard "UFS" type. It only changes how I/O is performed. How do I configure message queues?

Most Unix operating systems have message queue support by default. One way to check is to see if you have an However, you will likely need to increase the message queue parameters for Squid. Message queue implementations normally have the following parameters:

The messages between Squid and diskd are 32 bytes for 32-bit CPUs and 40 bytes for 64-bit CPUs. Thus, MSGSSZ should be 32 or greater. You may want to set it to a larger value, just to be safe.

We'll have two queues for each I've found that 75 messages per queue is about the limit of decent performance. If each diskd message consists of just one segment (depending on your value of MSGSSZ), then MSGSEG should be greater than 75.

MSGMNB and MSGTQL affect how many messages can be in the queues at one time. Diskd messages shouldn't be more than 40 bytes, but let's use 64 bytes to be safe. MSGMNB should be at least 64*75. I recommend rounding up to the nearest power of two, or 8192.

MSGTQL should be at least 75 times the number of FreeBSD

Your kernel must have options SYSVMSG

You can set the parameters in the kernel as follows. This is just an example. Make sure the values are appropriate for your system: options MSGMNB=8192 # max # of bytes in a queue options MSGMNI=40 # number of message queue identifiers options MSGSEG=512 # number of message segments per queue options MSGSSZ=64 # size of a message segment options MSGTQL=2048 # max messages in system OpenBSD

You can set the parameters in the kernel as follows. This is just an example. Make sure the values are appropriate for your system: option MSGMNB=16384 # max characters per message queue option MSGMNI=40 # max number of message queue identifiers option MSGSEG=2048 # max number of message segments in the system option MSGSSZ=64 # size of a message segment (Must be 2^N) option MSGTQL=1024 # max amount of messages in the system Digital Unix

Message queue support seems to be in the kernel by default. Setting the options is as follows: options MSGMNB="8192" # max # bytes on queue options MSGMNI="40" # # of message queue identifiers options MSGMAX="2048" # max message size options MSGTQL="2048" # # of system message headers


If you have a newer version (DU64), then you can probably use # sysconfig -q ipc To change them make a file like this called ipc.stanza: ipc: msg-max = 2048 msg-mni = 40 msg-tql = 2048 msg-mnb = 8192 then run # sysconfigdb -a -f ipc.stanza You have to reboot for the change to take effect. Linux

Stefan Köpsell reports that if you compile sysctl support into your kernel, then you can change the following values: kernel.msgmnb kernel.msgmni kernel.msgmax

Winfried Truemper reports: The default values should be large enough for most common cases. You can modify the message queue configuration by writing to these files: /proc/sys/kernel/msgmax /proc/sys/kernel/msgmnb /proc/sys/kernel/msgmni Solaris

Refer to in Sunworld Magazine.

I don't think the above article really tells you how to set the parameters. You do it in /etc/system with lines like this: set msgsys:msginfo_msgmax=2048 set msgsys:msginfo_msgmnb=8192 set msgsys:msginfo_msgmni=40 set msgsys:msginfo_msgssz=64 set msgsys:msginfo_msgtql=2048

Of course, you must reboot whenever you modify /etc/system before changes take effect. How do I configure shared memory?

Shared memory uses a set of parameters similar to the ones for message queues. The Squid DISKD implementation uses one shared memory area for each cache_dir. Each shared memory area is about 800 kilobytes in size. You may need to modify your system's shared memory parameters:

For Squid and DISKD, FreeBSD

Your kernel must have options SYSVSHM

You can set the parameters in the kernel as follows. This is just an example. Make sure the values are appropriate for your system: options SHMSEG=16 # max shared mem id's per process options SHMMNI=32 # max shared mem id's per system options SHMMAX=2097152 # max shared memory segment size (bytes) options SHMALL=4096 # max amount of shared memory (pages) OpenBSD

OpenBSD is similar to FreeBSD, except you must use option SHMSEG=16 # max shared mem id's per process option SHMMNI=32 # max shared mem id's per system option SHMMAX=2048 # max shared memory segment size (pages) option SHMALL=4096 # max amount of shared memory (pages) Digital Unix

Message queue support seems to be in the kernel by default. Setting the options is as follows: options SHMSEG="16" # max shared mem id's per process options SHMMNI="32" # max shared mem id's per system options SHMMAX="2097152" # max shared memory segment size (bytes) options SHMALL=4096 # max amount of shared memory (pages)


If you have a newer version (DU64), then you can probably use # sysconfig -q ipc To change them make a file like this called ipc.stanza: ipc: shm-seg = 16 shm-mni = 32 shm-max = 2097152 shm-all = 4096 then run # sysconfigdb -a -f ipc.stanza You have to reboot for the change to take effect. Linux

Winfried Truemper reports: The default values should be large enough for most common cases. You can modify the shared memory configuration by writing to these files: /proc/sys/kernel/shmall /proc/sys/kernel/shmmax /proc/sys/kernel/shmmni /proc/sys/kernel/shm-use-bigpages

Stefan Köpsell reports that if you compile sysctl support into your kernel, then you can change the following values: kernel.shmall kernel.shmmni kernel.shmmax Solaris

Refer to in Sunworld Magazine.

To set the values, you can put these lines in /etc/system: set shmsys:shminfo_shmmax=2097152 set shmsys:shminfo_shmmni=32 set shmsys:shminfo_shmseg=16 Sometimes shared memory and message queues aren't released when Squid exits.

Yes, this is a little problem sometimes. Seems like the operating system gets confused and doesn't always release shared memory and message queue resources when processes exit, especially if they exit abnormally. To fix it you can ``manually'' clear the resources with the ipcs | grep '^[mq]' | awk '{printf "ipcrm -%s %s\n", $1, $2}' | /bin/sh What are the Q1 and Q2 parameters?

In the source code, these are called cache_dir diskd /cache1 1024 16 256 Q1=72 Q2=64

If there are more than Q1 messages outstanding, then Squid will intentionally fail to open disk files for reading and writing. This is a load-shedding mechanism. If your cache gets really really busy and the disks can not keep up, Squid bypasses the disks until the load goes down again.

If there are more than Q2 messages outstanding, then the main Squid process ``blocks'' for a little bit until the diskd process services some of the messages and sends back some replies.

Q1 should be larger than Q2. You want Squid to get to the ``blocking'' condition before it gets to the ``refuse to open files'' condition.

Reasonable values for Q1 and Q2 are 72 and 64, respectively. Authentication How does Proxy Authentication work in Squid?

Note: The information here is current for version 2.5.

Users will be authenticated if squid is configured to use Browsers send the user's authentication credentials in the If Squid gets a request and the If the header is missing, Squid returns an HTTP reply with status 407 (Proxy Authentication Required). The user agent (browser) receives the 407 reply and then prompts the user to enter a name and password. The name and password are encoded, and sent in the NOTE: The name and password are encoded using ``base64'' (See section 11.1 of ). However, base64 is a binary-to-text encoding only, it does NOT encrypt the information it encodes. This means that the username and password are essentially ``cleartext'' between the browser and the proxy. Therefore, you probably should not use the same username and password that you would use for your account login.

Authentication is actually performed outside of main Squid process. When Squid starts, it spawns a number of authentication subprocesses. These processes read usernames and passwords on stdin, and reply with "OK" or "ERR" on stdout. This technique allows you to use a number of different authentication schemes, although currently you can only use one scheme at a time.

The Squid source code comes with a few authentcation processes for Basic authentication. These include: LDAP: Uses the Lightweight Directory Access Protocol NCSA: Uses an NCSA-style username and password file. MSNT: Uses a Windows NT authentication domain. PAM: Uses the Linux Pluggable Authentication Modules scheme. SMB: Uses a SMB server like Windows NT or Samba. getpwam: Uses the old-fashioned Unix password file. sasl: Uses SALS libraries. winbind: Uses Samba authenticate in a Windows NT domain

In addition Squid also supports the NTLM and Digest authentication schemes which both provide more secure authentication methods where the password is not exchanged in plain text. Each scheme have their own set of helpers and auth_param settings. You can not mix helpers between the different authentication schemes. For information on how to set up NTLM authentication see .

In order to authenticate users, you need to compile and install one of the supplied authentication modules found in the helpers/basic_auth/ directory, one of , or supply your own.

You tell Squid which authentication program to use with the auth_param basic program /usr/local/squid/bin/ncsa_auth /usr/local/squid/etc/passwd How do I use authentication in access controls?

Make sure that your authentication program is installed and working correctly. You can test it by hand.

Add some acl foo proxy_auth REQUIRED acl all src 0/0 http_access allow foo http_access deny all The REQURIED term means that any authenticated user will match the ACL named Squid allows you to provide fine-grained controls by specifying individual user names. For example: acl foo proxy_auth REQUIRED acl bar proxy_auth lisa sarah frank joe acl daytime time 08:00-17:00 acl all src 0/0 http_access allow bar http_access allow foo daytime http_access deny all In this example, users named lisa, sarah, joe, and frank are allowed to use the proxy at all times. Other users are allowed only during daytime hours. Does Squid cache authentication lookups?

Yes. Successful authentication lookups are cached for one hour by default. That means (in the worst case) its possible for someone to keep using your cache up to an hour after he has been removed from the authentication database.

You can control the expiration time with the Note: This has nothing to do with how often the user needs to re-authenticate himself. It is the browser who maintains the session, and re-authentication is a business between the user and his browser, not the browser and Squid. The browser authenticates on behalf of the user on every request sent to Squid. What this parameter controls is only how often Squid will ask the defined helper if the password is still valid. Are passwords stored in clear text or encrypted?

Squid stores cleartext passwords in itsmemory cache.

Squid writes cleartext usernames and passwords when talking to the external authentication processes. Note, however, that this interprocess communication occors over TCP connections bound to the loopback interface or private UNIX pipes. Thus, its not possile for processes on other comuters or local users without root privileges to "snoop" on the authentication traffic.

Each authentication program must select its own scheme for persistent storage of passwords and usernames. How do I use the Winbind authenticators?


Winbind is a recent addition to Samba providing some impressive capabilities for NT based user accounts. From Squid's perspective winbind provides a robust and efficient engine for both basic and NTLM challenge/response authentication against an NT domain controller.

The winbind authenticators have been used successfully under Linux, FreeBSD and Solaris.

Supported Samba Releases

Squid 2.5 uses an internal Samba interface to communicate with the winbindd daemon. It is therefore sensitive to any changes the Samba team may make to the interface.

The winbind helpers shipped with Squid-2.5.STABLE2 supports Samba-2.2.6 to Samba-2.2.7a and hopefully later Samba-2.X versions. To use Squid-2.5.STABLE2 with Samba versions 2.2.5 or ealier the new --with-samba-sources=... configure option is required. This may also be the case with Samba-2.2.X versions later than 2.2.7a or if you have applied any winbind related patches to your Samba tree.

Squid-2.5.STABLE1 supported Samba 2.2.4 or 2.2.5 only. Use of Squid-2.5.STABLE2 or later recommended with current Samba-2.X releases.

For Samba-3.X the winbind helpers shipped with Squid should not be used (and won't work if your attempt to do so), instead the ntlm_auth helper shipped as part of the Samba-3 distribution should be used. This helper supports all versions of Squid and both the ntlm and basic authentication schemes. For details on how to use this Samba helper see the Samba documentation. For group membership lookups the wbinfo_group helper shipped with Squid can be used (this is just a wrapper around the samba wbinfo program and works with all versions of Samba) Configure Samba

Build/Install Samba

Samba must be built with these configure options: --with-winbind --with-winbind-auth-challenge (needed for ntlm)

Optionally, if building Samba 2.2.5, apply the patch. See below to determine if the patch is worthwhile. Test Samba's winbindd Edit smb.conf for winbindd functionality. The following entries in the [global] section of smbd.conf may be used as a template. workgroup = mydomain password server = myPDC security = domain winbind uid = 10000-20000 winbind gid = 10000-20000 winbind use default domain = yes Join the NT domain as outlined in the winbindd man page for your version of samba. Test winbindd functionality. Start nmbd (required to insure proper operation). Start winbindd. Test basic winbindd functionality "wbinfo -t": # wbinfo -t Secret is good Test winbindd user authentication: # wbinfo -a mydomain\\myuser%mypasswd plaintext password authentication succeeded error code was NT_STATUS_OK (0x0) challenge/response password authentication succeeded error code was NT_STATUS_OK (0x0) SMBD and Machine Trust Accounts

Samba 2.2.x

Samba's smbd daemon, while not strictly required by winbindd may be needed to manage the machine's trust account.

Well behaved domain members change the account password on a regular basis. Windows and Samba servers default to changing this password every seven days.

The Samba component responsible for managing the trust account password is smbd. Smbd needs to receive requests to trigger the password change. If the machine will be used for file and print services, then just running smbd to serve routine requests should keep everything happy.

However, in cases where Squid's winbind helpers are the only reason Samba components are running, smbd may sit idle. Indeed, there may be no other reason to run smbd at all.

There are two sample options to change the trust account. Either may be scheduled daily via a cron job to change the trust password.

is a sample perl script to load smbd, connect to a Samba share using smbclient, and generate enough dummy activity to trigger smbd's machine trust account password change code.

is a patch to Samba 2.2.5's smbpasswd utility to allow changing the machine account password at will. It is a minimal patch simply exposing a command line interface to an existing Samba function.

Note: This patch has been included in Samba as of 2.2.6pre2.

Once patched, the smbpasswd syntax to change the password is: smbpasswd -t DOMAIN -r PDC Samba 3.x

The Samba team has incorporated functionality to change the machine trust account password in the new "net" command. A simple daily cron job scheduling "net rpc changetrustpw" is all that is needed.

Configure Squid

Build/Install Squid Samba-2.X Squid must be built with the configure options: --enable-auth="ntlm,basic" --enable-basic-auth-helpers="winbind" --enable-ntlm-auth-helpers="winbind" --enable-external-acl-helpers="wb_group" Test Squid without auth

Before going further, test basic Squid functionality. Make sure squid is functioning without requiring authorization.

Test the helpers

Testing the winbind ntlm helper is not really possible from the command line, but the winbind basic authenticator can be tested like any other basic helper: # /usr/local/squid/libexec/wb_auth -d /wb_auth[65180](wb_basic_auth.c:136): basic winbindd auth helper ... mydomain\myuser mypasswd /wb_auth[65180](wb_basic_auth.c:107): Got 'mydomain\myuser mypasswd' from squid (length: 24). /wb_auth[65180](wb_basic_auth.c:54): winbindd result: 0 /wb_auth[65180](wb_basic_auth.c:57): sending 'OK' to squid OK The helper should return "OK" if given a valid username/password.


As Samba-3.x has it's own authentication helper there is no need to build any of the Squid authentication helpers for use with Samba-3.x. You do however need to enable support for the ntlm scheme if you plan on using this. Also you may want to use the wbinfo_group helper for group lookups --enable-auth="ntlm,basic" --enable-external-acl-helpers="wbinfo_group" Edit squid.conf

Setup the authenticators.

Add the following to enable both the winbind basic and ntlm authenticators. IE will use ntlm and everything else basic: auth_param ntlm program /usr/local/squid/libexec/wb_ntlmauth auth_param ntlm children 5 auth_param ntlm max_challenge_reuses 0 auth_param ntlm max_challenge_lifetime 2 minutes auth_param basic program /usr/local/squid/libexec/wb_auth auth_param basic children 5 auth_param basic realm Squid proxy-caching web server auth_param basic credentialsttl 2 hours

Note: For Samba-3.X the Samba ntlm_auth helper is used instead of the wb_ntlmauth and wb_auth helpers above. This helper supports all Squid versions and both ntlm and basic schemes via the --helper-protocol= option. See the Samba documentation for details. Add acl entries to require authentication: acl AuthorizedUsers proxy_auth REQUIRED .. http_access allow all AuthorizedUsers

Test Squid with auth

Internet Explorer:

Test browsing through squid with IE. If logged into the domain, a password prompt should NOT pop up.

Confirm the traffic really is being authorized by tailing access.log. The domain\username should be present.

Netscape, mozilla, opera...:

Test with a non-IE browser. A standard password dialog should appear.

Entering the domain should not be required if the user is in the default domain and "winbind use default domain = yes" is set in smb.conf. Otherwise, the username must be entered in "domain\username" format.

If no usernames appear in access.log and/or no password dialogs appear in either browser, then the acl/http_access portions of squid.conf are not correct.


Terms and Definitions Neighbor

In Squid, In Harvest 1.4, neighbor referred to what Squid calls a sibling. That is, Harvest had Regular Expression

Regular expressions are patterns that used for matching sequences of characters in text. For more information, see and . Security Concerns Open-access proxies

Squid's default configuration file denies all client requests. It is the administrator's responsibility to configure Squid to allow access only to trusted hosts and/or users.

If your proxy allows access from untrusted hosts or users, you can be sure that people will find and abuse your service. Some people will use your proxy to make their browsing anonymous. Others will intentionally use your proxy for transactions that may be illegal (such as credit card fraud). A number of web sites exist simply to provide the world with a list of open-access HTTP proxies. You don't want to end up on this list.

Be sure to carefully design your access control scheme. You should also check it from time to time to make sure that it works as you expect. Mail relaying

SMTP and HTTP are rather similar in design. This, unfortunately, may allow someone to relay an email message through your HTTP proxy. To prevent this, you must make sure that your proxy denies HTTP requests to port 25, the SMTP port.

Squid is configured this way by default. The default http_access deny !Safe_ports (additional http_access lines ...)

Do NOT add port 25 to ). You may want to make a cron job that regularly verifies that your proxy blocks access to port 25. $Id: FAQ.sgml,v 1.215 2004/02/05 17:06:11 wessels Exp $