Page 2 of 4 FirstFirst 1234 LastLast
Results 11 to 20 of 37

Thread: High Java CPU usage question

  1. #11
    Join Date
    Nov 2008
    Posts
    119
    Rep Power
    7

    Default

    I am having this problem as well.
    100% dual cpu consumption. The whole system is hung.
    It consistently happens every 24 to 48 Hrs.

    Ubuntu 10.04 x64
    Zimbra 7.1.3 w/patch 7.1.3

    I have to hard reset/crash the system to get zimbra running again
    Guest Ubuntu 12.04LTS x64 and Zimbra 8.0.5
    Host CentOS 6.4 x64 :: KVM :: LVM :: Hardware RAID

  2. #12
    Join Date
    Nov 2008
    Posts
    119
    Rep Power
    7

    Default

    my last zimbra directory backup was 8.6 GB (disk to disk copy)
    - dual 2.2 Ghz processors
    - 100 GB HDD
    - 8 GB RAM

    the system runs very low cpu, consistently.

    there is a bug somewhere causing 100% cpu loop, an unresponsive system
    Guest Ubuntu 12.04LTS x64 and Zimbra 8.0.5
    Host CentOS 6.4 x64 :: KVM :: LVM :: Hardware RAID

  3. #13
    Join Date
    Nov 2009
    Location
    Ljubljana, Slovenia
    Posts
    268
    Rep Power
    6

    Default

    Mine ZCS 7.1.2 also! Started this week - it was all fine until this week. No changes made to ZCS server for a month or two for sure.

    QuadXEON, 3,2 GHz
    8 GB RAM
    Ubuntu 8.04 LTS, 64-bit
    ZCS 7.1.2
    Domains: 300+
    Users: 2000+

    Before: CPU average 12%
    This week: CPU average: 64%
    CPU hog process: JAVA

    ** EDIT **
    I upgraded to 7.1.4 ZCS and CPU went back to normal. Actually, it went way down below previous average.
    Now my ZCS is using on average 3% CPU, which is more than satisfactory!
    Last edited by Labsy; 01-19-2012 at 03:25 PM.

  4. #14
    Join Date
    Oct 2007
    Posts
    57
    Rep Power
    8

    Default

    I still have the problem with 7.1.4. Did you used some extra commands to boost performance and delay certain automated checks?

    Grtz,
    Roderick
    FC11 x64 / Centos 6.2 - Zimbra 7.2.0 Open Source version - ESX 5.0

  5. #15
    Join Date
    Nov 2009
    Location
    Ljubljana, Slovenia
    Posts
    268
    Rep Power
    6

    Default

    No, nothing special was configured, just upgraded to 7.1.4.

    But in my case I had one separate problem on ESX host, which begun using ESX swap due to overcommited memory settings. Memory swapping on ESX is a huge performance impact, which might also be guilty in part for my performance problems with ZCS.

    As I see you are on ESX 4.1, so check in vSphere client --> ESX host --> Performance --> Default or Disk --> enable "Swap-in" and "Swap-out" graphs. If either of those is not zero, then you ESX host is swapping memory to disk, which is very, very bad. If this is your case, see the largest swap consumption (mine was 500 MB) and lower RAM MAX for all VMs so, that total sum of max RAM usage would be lower for at least swap max consumption (in my case, I reduced RAM MAX values for 550 MB).
    After few days I checked "Swap-in" And "Swap-out" graphs and they were both at ZERO, my server's performance is fine since then
    Last edited by Labsy; 02-12-2012 at 05:30 AM.

  6. #16
    Join Date
    Oct 2007
    Posts
    57
    Rep Power
    8

    Default

    Thnx for the reply!
    There is no ESX swapping, enough free memory available. And the thing is the high spikes also occur after installing a new guest OS and Zimbra, without any mailboxes and users.

    It works though but the cause of the spikes is a java process that is executed every 2 to 2:20 minutes. I wonder what this process is and how it can be rescheduled to a longer time interval.

    You can see graphs and more info here:
    http://www.zimbra.com/forums/adminis...c-7-1-4-a.html

    Grtz,
    Roderick
    FC11 x64 / Centos 6.2 - Zimbra 7.2.0 Open Source version - ESX 5.0

  7. #17
    Join Date
    Sep 2008
    Posts
    98
    Rep Power
    7

    Default

    Quote Originally Posted by mrcoffee11 View Post
    Thnx for the reply!
    There is no ESX swapping, enough free memory available. And the thing is the high spikes also occur after installing a new guest OS and Zimbra, without any mailboxes and users.

    It works though but the cause of the spikes is a java process that is executed every 2 to 2:20 minutes. I wonder what this process is and how it can be rescheduled to a longer time interval.

    You can see graphs and more info here:
    http://www.zimbra.com/forums/adminis...c-7-1-4-a.html

    Grtz,
    Roderick
    hi,

    did you found a solution? or anybody else
    i have the same problem after upgrade my system to 7.1.4
    my system getting very slow after some hours uptime.
    currently 32 java processes are running: ps aux | grep -i java | wc -l

    used version:
    esxi environment: esxi 5.0.0
    os: centos 6 2.6.32-71.el6.x86_64
    zimbra: 7.1.4_GA_2555.RHEL6_64_20120105094542 CentOS6_64 FOSS edition

    greetings
    gregor

  8. #18
    Join Date
    Oct 2007
    Posts
    57
    Rep Power
    8

    Default

    Sadly not... See this thread which I started around two weeks ago.
    http://www.zimbra.com/forums/newrepl...ote=1&p=233240
    FC11 x64 / Centos 6.2 - Zimbra 7.2.0 Open Source version - ESX 5.0

  9. #19
    Join Date
    Nov 2008
    Posts
    119
    Rep Power
    7

    Default My high CPU :: 100% CPU Consumption Problem is Resolved

    As it turns out, this is not a Zimbra problem but an environment problem.

    No logs to help debug this problem so lots of digging and "the process of elimination". It has taken me the past three weeks just to verify stability.

    Anyhow, to make a long story short, since this is an environment problem with my guest VM ubuntu 10.04 Zimbra server related to native I/O on LVM.

    I resolved my Problem by turning off/not enabling ACPI and turning off/not enabling ACIP in my guest VM ubuntu 10.04 Zimbra server (7.1.3 w/patch)
    Guest Ubuntu 12.04LTS x64 and Zimbra 8.0.5
    Host CentOS 6.4 x64 :: KVM :: LVM :: Hardware RAID

  10. #20
    Join Date
    Oct 2007
    Posts
    57
    Rep Power
    8

    Default

    Thnx for sharing your experience.
    Doesn't work for me though, I don't got ACPI installed on my VM. When I shut down Zimbra the CPU spikes are not there anymore.
    FC11 x64 / Centos 6.2 - Zimbra 7.2.0 Open Source version - ESX 5.0

Similar Threads

  1. [SOLVED] High cpu load, only on one core?.. java..
    By ducommun in forum Administrators
    Replies: 1
    Last Post: 02-25-2011, 06:40 AM
  2. High CPU usage on server and client
    By jack-Z in forum Administrators
    Replies: 10
    Last Post: 01-07-2009, 10:12 PM
  3. Replies: 5
    Last Post: 02-04-2008, 05:55 PM
  4. HIGH CPU use
    By brwatters in forum Administrators
    Replies: 3
    Last Post: 07-05-2006, 09:47 AM
  5. 3.1 on FC4 problems
    By cohnhead in forum Installation
    Replies: 8
    Last Post: 05-26-2006, 12:16 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •