dying children?

[Jon Dustin]

Greetings -

For the past few weeks my Mailscanner duo has been slowing down. I have a pair of SLES10 boxes running Postfix and MailScanner v4.62.9. I realize this is a slightly old version, but it had been running very well for a long time.

Last night I began digging around and discovered the MailScanner children are dying very quickly (just a few minutes), and replaced with new siblings. If I understand MailScanner's architecture, the children should live for 12 hours before being replaced by the master?

This morning I have upgraded one MailScanner node to the latest STABLE version (v4.70.7), but the problem still remains. 

I thought the issue may have to do with MailWatch, so I disabled the "Always Looked Up Last" function. This change had no effect.

At this point it appears some messages are being scanned multiple times by different children, which slows the process down dramatically. Thinking this was a locking issue, I changed Lock Type to posix, also with no effect.

The speed reported in syslog seems decent for most batches:

Jul 29 10:31:31 mail2 MailScanner[12348]: Batch (30 messages) processed in 229.92 seconds
Jul 29 10:31:39 mail2 MailScanner[12601]: Batch (13 messages) processed in 107.58 seconds
Jul 29 10:32:02 mail2 MailScanner[12375]: Batch (22 messages) processed in 185.64 seconds
Jul 29 10:32:07 mail2 MailScanner[12685]: Batch (12 messages) processed in 99.20 seconds
Jul 29 10:32:25 mail2 MailScanner[12685]: Batch (2 messages) processed in 17.73 seconds
Jul 29 10:32:26 mail2 MailScanner[12804]: Batch (4 messages) processed in 44.68 seconds

This box is running as a VM session, but on very decent hardware with 1GB ram.

Any thoughts on where I should investigate next?

Thank you for your assistance and ideas.