[SOLVED] Zimbra Stops Responding and Im Out Of Ideas

[ndconsulting]

Hi everyone. I posted last week sometime with the same issue with no luck.

Ive got a new zimbra install that will not run more than 8 hours at all. The length of time is stays active seems to be dependent on user load, but overall after ~8 hours, IMAP and Webmail become unresponsive. The only way to resolve it seems to do a zmcontrol stop / start.

Its a debian 5 box with 4gb ram with everything up to date as of 9/1. Kernel Version 2.6.24-24-xen

There isin't much load on the box. There are only 35 users checking via IMAP.

My biggest problem is that I have read through the documentation, but I really don't know where to start. I have gone through all the log files and nothing seems to stand out around the time it locks up. But Im new to zimbra so im not exactly sure what to look for specifically.

Can anyone provide me some direction? About another day or so and Im going to have to find another mail server.

[bdial]

is this a virtual machine?

[ndconsulting]

Yes it is a Xen VM running at Slicehost - com. Its one of their 4gb slices.

When IMAP and webmail stop responding, killing the java process will bring back up webmail, but not imap. I did a kill -QUIT xxx on the java process one of the past few times and ive attached the stacktrace. Not sure if that helps at all.

When it stops, memory usage seems decent. Overall shows almost 800MB free and the server itself still responds no problem. SSH session works fine and it seems to be running fast.

[ndconsulting]

Finally solved the issue so I figured I would post what I did to resolve it.

I took two actions that solved the problem. 6.0.1 was released, and it mentioned some bug fixes for Debian Lenny. When I went to run a backup of the /zimbra directory I noticed it was taking much longer than it should. Turns out, the daily mailbox.log files were gigantic, averaging 12-15gigs each! Since the server was on a VPS server anyway and disk IO probably isin't that stellar to being with, im sure writing a log file that large over a 8-10 hour business day was totally destroying what mediocre IO performance did exist.

After turning logging down to WARN for imap, the log file is now averaging 250k a day and the server has been up for over 24 hours (thats never happened).

Second thing I did was to go ahead and run the 6.0.1 update as well.

Not sure which one cured it, but at least its running reliably.