Koozali.org: home of the SME Server

Obsolete Releases => SME Server 9.x => Topic started by: mmccarn on May 19, 2014, 07:35:48 PM

Title: SME 9 RC1 Server Load
Post by: mmccarn on May 19, 2014, 07:35:48 PM
[informational]

After upgrading from SME9beta to SME9RC1, my qpsmtpd quickly started reporting "Too many connections".

I have found (so far) that the only way to keep the connection count at a reasonable level has been to disable clamd and reboot (the email-update event didn't seem to force qpsmtpd to re-read the updated /var/service/qpsmtpd/config/peers/0 file)
Code: [Select]
config setprop clamd status disabled
signal-event email-update
signal-event reboot

With clamd disabled, my connection count never exceeds 10 - 11.  With clamd enabled, I was reaching 50+ connections within minutes of rebooting.

I'm pretty sure the cause is that I'm running SME as a spam filter on a virtual box that is under-resourced for the load, with bad disk performance on top.  I have another box that is doing virus scanning before anything reaches anyone's Inbox...

Title: Re: SME 9 RC1 Server Load
Post by: ReetP on May 20, 2014, 12:04:40 AM
If it doesn't work as expected - bug it :-)

If you can do a bug then the devs can help you figure out if there is a problem with the software or your hardware, and everyone gets the benefit !

B. Rgds
John
Title: Re: SME 9 RC1 Server Load
Post by: mmccarn on May 20, 2014, 01:38:38 PM
If it doesn't work as expected - bug it :-)
John

Thanks.

There is a bug (bug 8203 (http://bugs.contribs.org/show_bug.cgi?id=8203)) I opened on this when SME9 RC3 came out -- but I suspect that my problem is specific to my situation - somewhere between my virtualized SME running on an 8 year old server, too many other VMs on the same host, too much email (~15,000 per day), using an "internal mail server", and (possibly?) changes to clamav. 

I wanted to document the work-around I'm using in case anyone else runs into the same problem.
Title: Re: SME 9 RC1 Server Load
Post by: ReetP on May 20, 2014, 03:43:01 PM
Probably best to add that to the bug, and possibly reopen if you still have problems.