Hi All, I have been using several (identical) instances of James (3.0-beta3), in a large corporate environment, with some mailets I have written to perform some standard manipulations on messages as they pass to and from another MTA which actually scans the mail content
Each instance processes many thousands of emails per day and we have had very few issues, but there have been a couple of cases where users have reported long delays before delivery of their messages. Now there is pressure to fix the issue Detailed monitoring the traffic initially revealed that the delays are inversely related to load - so high load, low delay and vice-versa Later it was discovered that this behaviour appears to start randomly on some of the James instances - I've been told that rebooting the server always cures the problem (until the problem randomly starts again), and I can confirm that simply restarting James usually, but not always, fixes the problem Further investigation of the log files reveals the reason for the load-delay relationship - it appears that James enters a mode where it will only pass on messages when it receives a new incoming message - so high load, short interval between incoming messages and therefore short delays but at night and at weekends, there can be several minutes between incoming messages and we get correspondingly long delivery delays My problem at the moment is that I really don't know where to start looking for the source of problem so... - Is this by any chance a known issue? - Has anyone seen this behaviour before and if so, did you find a solution? - I would really appreciate any ideas of how to approach the problem Thanks in advance -- Phil
