Ed

I just installed Clamav snapshot from last night and I'm using the processes instead of threads and it's only got 3 clamd's running now. Seems to be just as fast as before with 6 clamd threads.

I don't think I'm Disk I/O bound. I'm using SDS to do RAID 1+O on a 2 SCSI controllers and a D1000 with 12x 9 gig SCSI drives. I can push 32Megabytes/sec writes. I can do even better with 2 more SCSI controllers and another D1000.

Right now this machine is grinding through a 140K email backlog. clamd died last night, normally I see 750K emails/day with loads between 4 and 8. It's only been up for 6 days. Gotta love the old SBus Sun Enterprise machines.

thanks
Ken

On Dec 2, 2003, at 1:48 PM, Ed Phillips wrote:

On Tue, 2 Dec 2003, Ken McKittrick wrote:

Ed

Amavisd-new and Clamd are the bottleneck on this system. With 8 amavisd
processes, clamd runs with 5 threads. That goes up to 6 threads with 12
amavisd processes. This might be the most I can get on an 8 cpu system.

We use MIMEDefang + clamd to scan about 300,000 emails per day on a V880
which is mostly idle.


I'm still fine tuning this machine for maximal throughput.

Are you using /var/tmp for the temporary files while scanning? My guess
is that your bottleneck might be disk i/o. We average about 2 active
clamd threads, I think, but I haven't watched it closely...


Ed


Thanks Ken McKittrick USADatanet

On Dec 2, 2003, at 10:20 AM, Ed Phillips wrote:

On Tue, 2 Dec 2003, Ken McKittrick wrote:

Hello

I have Postfix 2.0.16, Amavisd-new, SpamAssassin, and Clamav 0.65
running on a Enterprise 4500 with 8 CPU's. It's configured to use 16
Amavisd processess and the Clamav.conf has MaxThreads set to 40. I
only
see 6 clamd threads. I don't think that Clamd is keeping up with
Amavisd.

From looking at the source, clamd appears to allocate a thread when a
request comes in... and then the thread exits. It doesn't pre-allocate
MaxThreads threads and have them sit idle waiting for work...


Hope this helps!

Ed

Any ideas on how to get more Clamd threads?

25825 vscan 28M 23M run 20 0 0:00.10 3.6% amavisd/1
25957 vscan 28M 21M run 20 0 0:00.08 3.5% amavisd/1
25978 vscan 28M 21M run 20 0 0:00.07 3.0% amavisd/1
26030 vscan 28M 20M cpu8 30 0 0:00.06 2.8% amavisd/1
26022 vscan 28M 21M run 20 0 0:00.05 2.5% amavisd/1
26094 vscan 27M 20M run 20 0 0:00.04 2.3% amavisd/1
25984 vscan 27M 20M run 20 0 0:00.05 2.3% amavisd/1
26131 vscan 27M 20M sleep 20 0 0:00.04 2.0% amavisd/1
26182 vscan 27M 20M run 20 0 0:00.02 1.5% amavisd/1
12752 vscan 10M 9416K sleep 58 0 0:00.09 1.5% clamd/6
26247 vscan 27M 20M cpu13 20 0 0:00.02 1.2% amavisd/1
26252 vscan 26M 19M run 30 0 0:00.01 1.0% amavisd/1
26265 vscan 27M 19M cpu1 30 0 0:00.01 0.9% amavisd/1
26267 vscan 26M 19M run 20 0 0:00.01 0.8% amavisd/1


Thanks
Ken McKittrick
USADatanet




------------------------------------------------------- This SF.net email is sponsored by OSDN's Audience Survey. Help shape OSDN's sites and tell us what you think. Take this five minute survey and you could win a $250 Gift Certificate. http://www.wrgsurveys.com/2003/osdntech03.php?site=8 _______________________________________________ Clamav-users mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/clamav-users

Reply via email to