We have had some issues with zimbra hanging with high CPU load. A java process takes up 100% of one core on the box and then no one can access email either through imap or the web gui. The process just times out. Eventually after 10+ minutes zimbra will correct itself. Restarting zimbra will fix the problem right away. There doesn't appear to be any lag with ssh or other processes as the java process only ties up one core.
Viewing the logs show nothing of interest except the fact that zimbra still seems to be processing emails even though users are timing out.
The problem has occured in the middle of the night when there hasn't been a high load on the server(that is how I know it eventually fixes itself) and also during the day with normal load. When it happened during the day I wasn't thinking too much about gathering information but trying to fix the problem which ultimatily ended up in restarting zimbra. I did consider killing the java process but had no idea how that would effect zimbra.
While it would be great if someone knew what this could be I realize I am lacking infomration to help diagnose the problem. If anyone could offer ideas as to what I should do next time this happens then that would be helpfull. Ideas such as grab the process list(which I am kicking myself for not doing) or look more closely at certain logs or certain things in logs that I might be missing/glancing over. Any help would be gratefull.