High Java CPU usage question

[mek1]

Release 7.1.0_GA_3140.UBUNTU10_64 UBUNTU10_64 NETWORK edition.

OS: Ubuntu server 10.04 LTS x64

I just installed a trial of ZCS 7.1 NE for testing on a VMware ESXi 4.1 host. We noticed the server running unusually higher than our two production machines both with 100+ active accounts. The new 7.1 server has 4x 3.33 GHz processors and 4 GB RAM allocated with no active users accounts yet. Seeing as there is no one using the server yet I find this a bit odd. If we restart Zimbra or the server it goes back down to an acceptable level for a while then kicks back up to what we're seeing below.

Does anyone have performance recommendations for this scenario?


ZCS 7.1 Server stats:

CPU load averages 1.21 (1 min) 1.19 (5 mins) 1.12 (15 mins)

Memory

free
             total       used       free     shared    buffers     cached
Mem:       4059624    3773196     286428          0     162364    1446564
-/+ buffers/cache:    2164268    1895356
Swap:     11894776      74724   11820052

Uptime

 
09:13:38 up 3 days, 15:27,  2 users,  load average: 1.09, 1.16, 1.11

Htop process that is consistently running high:

PID USER     PRI  NI  VIRT   RES   SHR S CPU% MEM%   TIME+  Command                                                                       
11807 zimbra    20   0  654M  234M  5388 R 94.0  5.9 41h37:43 /opt/zimbra/java/bin/java -XX:ErrorFile=/opt/zimbra/log -client -Xmx256m -Dzim

[SkyHiRider]

Somethings definitely od with Java, as I ran into a similar problem today - both cpu's were spiking at 90%, had to restart Zimbra to fix it.

This is very bad, is there a solution for this?

[Naik]

Does anyone have any idea about this? I have the same problem and restart seems to fix it temporarily but it goes back to high CPU usage with spikes in 90s. What is using Java process that is using up so much CPU?

Your help is appreciated.

top command on mailbox VM

Cpu(s): 42.9%us, 35.6%sy, 0.0%ni, 21.3%id, 0.0%wa, 0.2%hi, 0.0%si, 0.0%st


PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND
4998 zimbra 18 0 1657m 975m 14m S 151.3 24.7 884:48.22 java
28314 root 15 0 282m 15m 8632 S 2.7 0.4 0:00.51 gnome-terminal
15973 root 15 0 97048 12m 6340 S 2.0 0.3 0:06.06 Xorg
28301 root 15 0 271m 14m 10m S 1.0 0.4 0:03.74 gnome-system-mo
4937 zimbra 15 0 1590m 337m 5656 S 0.3 8.5 2:33.67 mysqld
28158 root 15 0 257m 7340 5656 S 0.3 0.2 0:00.21 gnome-settings-

[phoenix]

Originally Posted by Naik

Does anyone have any idea about this?

Not unless somebody filed a bug report about the problem.

Originally Posted by Naik

I have the same problem and restart seems to fix it temporarily but it goes back to high CPU usage with spikes in 90s. Your help is appreciated.

I'd suggest you update to the latest 7.1.3 release of Zimbra and see if the problem still occurs, if it does then open a support case and/or file a bug report about the problem.

[Naik]

Originally Posted by phoenix

Not unless somebody filed a bug report about the problem.

I'd suggest you update to the latest 7.1.3 release of Zimbra and see if the problem still occurs, if it does then open a support case and/or file a bug report about the problem.

Thanks Phoenix for the quick reply but we are already on 7.1.2 and it diddnt resolve. We would like to know something about this before we upgrade. Do you know what on the mailbox server uses java?

[Naik]

So it is probably tomcat using Java. But still dont know why Java is going crazy. I think it is probably one of the bad coded zimlet but dont even know which one to start on and not even 100% if it is a zimlet... I am suspecting Social Zimlet...

[soba@ukw.edu.pl]

Update to 7.1.2 P1 I was the same problem. Update clear this situation.

[ThorGoLucky]

Upgraded to 7.1.3 and Zimbra is still very slow (several seconds to send a message) and java utilized almost half of the total memory.

[ulaaser]

Same problem here... i have installed latest version with patch inside centos 6.0 64bit in esxi 2vcpus and 6gb ram.

there are a growing number of java proceses all of them taking up all cpu ressources... memory is stable at 3gb

but after a few hours (4-6) the server becomes unresponsive and i must connect via console to restart the server...

i am also unable to kill the process /opt/zimbra/java/bin/java -version (varios of them are using the cpu ressources, and are not killed if im using zmcontrol restart or kill PID)

[ccelis5215]

Uhmmm. that's happened in my server where i've allocated 2 gigs.