High CPU usage, Massive slow down

[deoriger]

I've spent quite a bit of time searching the forums for an answer to our problems, but so far no luck. So it looks like it's time to post and see if someone might be able to help.

We are running 5.0.8NE on RHEL4.0. The machine has dual Xeon's with 6GB of RAM. This server supports 1500+ users with about 90-95% using the web interface.

The problem we are experiencing is a constant high CPU usage. Top shows 95%+ CPU usage with no idle. The main process swallowing up all this resource is java. The problem cropped up on Friday around 10:30am and we haven't been able to get it to go away. I used zmstat-chart to create graphs from the past few days. Graphs from earlier in the week show normal CPU usage, but on Friday there is a clear CPU ramp up on the graphs. This problem has continued, even through the weekend. Restarting, rebooting, even failing over to the second server doesn't resolve this problem. We've tried a few changes to the memory heap of java and mysql, but nothing seems to work.

Any ideas would be greatly appreciated. We might go to the backups to resolve the problem, but understanding what and why would help us sleep better at night.

[iway]

We have the same problem. 8GB Ram, Core2Dual processor, 700 users. CPU usage is high, RAM is completely used up, with Java showing 70%.

Any ideas?

5.0.9 NE on Ubuntu 8.0.4_64

[bdial]

did you check like the mailbox log for around htat time to see if anything might have happened? Have you tried to disable the logger for the time being too? It seems like a lot of these problems I've seen have been issues with the logger, like mysql database sizes or corruption or something.

[iway]

No, have not found anything unusual in mailbox.log.

How can I disable the logger safely? Does it have any negative effects on the system?

[bdial]

su - zimbra
zmloggerctl stop

the only thing it'll do is stop updating statistics in the web interface. Theres a few instances in the wiki that talk about problems like this such as http://wiki.zimbra.com/index.php?tit...ata_Corruption

[deoriger]

We actually fixed this awhile back. Thought I'd posted. Anyway, one of the other Admins upgraded the server from 5.0.8 to 5.0.10. During the process, the system checks fixed what ever was wrong. We still don't know what was actually broken, but what ever it was it's now resolved. Server's been running smoothly ever since.