Strange load situation

[Steve Campbell]

I have a strange problem going on with one of my mailservers. Recently, the 
load average has been climbing, and I can't seem to discover why. This 
morning, I decided to see if my purges for the Mailwatch databases were 
running properly, and started a purge. The load average climbed to above 14, 
and sendmail stopped accepting mail. Since nothing was being accepted, I 
decided to stop MailScanner to lighten the load and let the purge finish. I 
saw something I had not seen before after doing a 'ps' - Mailscanner was 
finishing batches. I know this is normal, but not usually seen. It stayed 
down for a few minutes.

After the purge finished, I restarted MS, and lo and behold, the load 
average is staying down around a normal 2. This is a pretty hefty server. A 
lot of incoming mail is queueing up, and the Mailscanner children are all 
handling the default 30 messages. It's not really gaining on the queued 
mail, but this could be current flooding, so I'm not real concerned yet.

bdc and clamscan always seemed to be at the top of 'top', with MailScanner 
children just below, or mixed in between, but now, I don't always even see 
them in the top 20.

Can anyone indicate they have seen similar results? I'm not real concerned 
yet, at least not until I see that it's not going to catch up. The purge 
seemed to have been working, but I thought maybe there was a little bit too 
much in the maillog table - turns out not so. This would almost indicate 
it's better to slow down my MS scan interval and let more messages per batch 
be scanned per child, or maybe lower my messages per batch.

Steve Campbell
campbell at cnpapers.com
Charleston Newspapers