Performance issues!

[cnicholson]

We're running professional 2.0, dual 3.4Ghz Xeon w/ ESX 3.02, ProxMox gets 2.5G RAM, 30GB HD and dual vCPUs. We're having a HORRIBLE performance problem. We're at 140k messages/day and the server is pegged (vCPUs and host itself) nearly 100% of the time. I've run the proxperf command and it looks like my disk perf is ROTTEN:

CPUBOGOMIPS: 12811.51
REGEX/SECOND: 104755
HD SIZE: 28.54 (/dev/sda2)
BUFFERED READS: 20.06 MB/sec
AVERAGE SEEK TIME: 6.38 ms
FSYNCS/SECOND: 320.65

Currently, I'm running it ALONE on my ESX host, but I can vmotion another one over there, access the SAME LUN on the SAN over the same path and get 90-120MB/sec to a VM on SAME vmfs volume.

I've turned off SMTP on the firewall to let the box breath a little, but it's acting like it can't keep up? I have bayesian, SPF, DCC, RAZOR, OCR all turned off, it actually forwards to ANOTHER spam filter, so it doesn't need to do recip validation or anything. I don't think I'm out of the ballpark on load, right? We have about 100 domains on the box.

What else can we do to troubleshoot this?!

Thanks,

 

[cnicholson]

UPDATE: If I clean boot, stop SMTP and wait for the queues to empty, I can get this up to 450 FSYNCS/SECOND. My proc is still 100% and my queue is getting longer and longer with no hope of ever processing my current load of inbound mail... ?! I need some suggestions on what to turn off (any other checks, how to disable clam?) My total mail flow is about 10G/week, so it's not astronomical either. I've tried the various other suggestions from the forum as well (limiting connections, disabling OCR and other services, NOT flushing the queue) and am out of time. On a related note, I installed a fresh copy of 2.0 on a DL380G3 w/ 2x 3.0Ghz Xeons, 4GB RAM, 2x36GB HDs, and 5i with BBWC and only got 246 FSYNCS/SECOND, so maybe my performance isn't that bad??! I just need to know what to throw at this problem.

Thanks for ny help you can provide!

 

[cnicholson]

Here's a quick example of what goes on. Proxmox takes the message in and sits on it for two hours.. running it's processor at 100%.

Received: from scrub.secantnet.net (localhost [127.0.0.1])
by scrub.secantnet.net (Proxmox) with ESMTP id 32ED584EF1
for <cnicholson@secantcorp.com>; Mon, 15 Oct 2007 23:41:56 -0400 (EDT)
Received: from www.proxmox.com (www.proxmox.com [87.230.14.152])
by scrub.secantnet.net (Proxmox) with ESMTP id D06A385023
for <cnicholson@secantcorp.com>; Mon, 15 Oct 2007 21:37:28 -0400 (EDT)
Received: by www.proxmox.com (Postfix, from userid 33)
id 6579AB82017A; Tue, 16 Oct 2007 02:34:43 +0200 (CEST

 

[tom]

We're running professional 2.0, dual 3.4Ghz Xeon w/ ESX 3.02, ProxMox gets 2.5G RAM, 30GB HD and dual vCPUs. We're having a HORRIBLE performance problem. We're at 140k messages/day and the server is pegged (vCPUs and host itself) nearly 100% of the time. I've run the proxperf command and it looks like my disk perf is ROTTEN:

CPUBOGOMIPS: 12811.51
REGEX/SECOND: 104755
HD SIZE: 28.54 (/dev/sda2)
BUFFERED READS: 20.06 MB/sec
AVERAGE SEEK TIME: 6.38 ms
FSYNCS/SECOND: 320.65

Currently, I'm running it ALONE on my ESX host, but I can vmotion another one over there, access the SAME LUN on the SAN over the same path and get 90-120MB/sec to a VM on SAME vmfs volume.

I've turned off SMTP on the firewall to let the box breath a little, but it's acting like it can't keep up? I have bayesian, SPF, DCC, RAZOR, OCR all turned off, it actually forwards to ANOTHER spam filter, so it doesn't need to do recip validation or anything. I don't think I'm out of the ballpark on load, right? We have about 100 domains on the box.

What else can we do to troubleshoot this?!

Thanks,

Hi,
the results of proxperf are quite bad, so there is something strange here. is there a possibity to get ssh access for us to see details? just open ssh for our IP 213.129.239.114 and send the login credentials to support@proxmox.com.