YA Squid Thread - tweaking for performance

[pete]

I have squid set up and authenticating via ldap, apart from a few basic things (acls, auth and a couple of url rewrites) the config file is pretty vanilla.

 

Would anyone mind sharing a mature squid.conf? As complex as you like. I'm interested in getting the best use of available memory more than anything.

[Geoff]

Have a read of this:

 

http://www.linux-faqs.com/squid.php

[pete]

Cheers Geoff

 

Second question

 

I'm building Dansguardian [2.9.8.2] and enabling clamav and clamd at compile time. Unfortunately Clamav 0.90.1 (0.90 onwards) breaks the av plugin. Make errors out on line 275 of sourcedir/src/contentscanners/clamav.cpp complaining

contentscanners/clamav.cpp: In member function ‘virtual int clamavinstance::init(void*)’:
contentscanners/clamav.cpp:275: error: ‘cl_loaddbdir’ was not declared in this scope

 

A bit of googling* reveals the beginnings of a possible fix - has anyone else run into this?

 

*http://www.mail-archive.com/[email protected]/msg308977.html

 

edit: code tags

[Geoff]

I'm not using Dansguardian to do AV scanning. I'm relying on the desktop AV for that.

[tom_newton]

We are using Clam in the commercial filter - and thats running the latest DG, ergo, it must be possible. Wether we are scanning out of DG or clam I dunno - I will see if I can remember to ask next time I collar one of the devs. Failing that, ask on the mailing list - Phil reads that, and if anyone knows the answer, he will

[pete]

@Geoff

I like redundancy and 2 sets of av defs.

 

Asked on the list and got a reply from Phil - the fix will be in the next beta and clamdscan will suffice for now.

 

The clamav plugin is currently incompatible with ClamAV 0.90, as the API

has changed relative to the older 0.88 releases. I intend to fix this in

the next beta, but for now, use the clamdscan plugin instead - this

talks to ClamD running as an external process, so library

incompatibilities aren't an issue. AFAIK this should work fine.

[Geoff]

Using clamdscan is a better solution anyway. It scales better.

[tom_newton]

@Geoff - any particular reason for that? I know one advantage of the dg variant is "AV progress bars" - not sure why you believe it will scale less well...

[tom_newton]

Also... I should go beat Phil for replying on dg list in work hours... oh wait.. i'm posting on edugeek. Sshh, dont tell t'boss like

[Geoff]

When you use clamscan you invoke a new copy of clamscan for each thing you scan. When you use clamdscan, you communicate with clamd via the UNIX pipe and tell one copy of clamd to scan several things.

 

With many concurrent users, clamscan tends to eat a lot of processing time/memory. Clamdscan does not.