The Community Forums

Interact with an entire community of cPanel & WHM users!
  1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Tracking 550MB memory spike

Discussion in 'General Discussion' started by cfmorley, Oct 1, 2011.

  1. cfmorley

    cfmorley Registered

    Joined:
    Oct 1, 2011
    Messages:
    2
    Likes Received:
    0
    Trophy Points:
    1
    cPanel Access Level:
    Root Administrator
    Hi,

    I have a VPS (goDaddy) running cPanel WHM 11.30.3 (build 5) on Centos 5.6 (2.6.18-028stab091.2). WHM is set to autoupdate to RELEASE settings. I have 1GB of dedicated memory with 60-65% free most of the time.

    On the 27th I ran easyapache to build apache 2.2.21 no problems.
    On the 29th I ran it again because I needed another PHP module but it wouldn't complete with memory alloc errors.

    I have tracked this down to some process/driver generating a spike of memory request for 550Mb - any program or script trying to run then fails due to memory alloc problems (perl scripts, YUM etc).

    I killed _all_ the processes except (init, sshd & my shell) and this spike still happens.

    The spike can be seen using "cat /proc/user_beancounters" in the privvmpages but it is so narrow it is never seen held only ever maxheld. It happens every 1 minute _exactly_ (and yes I did kill crond). In fact maxheld privvm excedes the barrier and limit (266000 maxheld, 261120 limit)! "vmstat 1" occasioanlly core dumps on the minute tick! :S

    This still happens after reboot through WHM (or shell reboot) & container reboot.

    I wrote a C program to allocate and commit 500MB of RAM and on the minute tick I get exactly 1 more errorcnt for privvmpages.

    I don't see anything in the logs like error messages from a driver restarting or such.

    This alloc spike is causing all sorts of problems - upcp can be interrupted in the middle for example.

    Does anyone have _any_ ideas?

    (GoDaddy support think this is not a VM problem but I am not convinced... it broke by 'magic' between the 27th & 29th!)

    Thanks,

    Chris
     
  2. cfmorley

    cfmorley Registered

    Joined:
    Oct 1, 2011
    Messages:
    2
    Likes Received:
    0
    Trophy Points:
    1
    cPanel Access Level:
    Root Administrator
    I think this is a Virtuozzo/host kernel/module issue. After reprovision (server build on new VM from scratch - very inconvenient!) by GoDaddy the virtuozzo memory limits were altered (although they said several times they were correct before) and now the system does not throw malloc fails.

    The memory spike is still there 168,000 pages.

    I don't think it is a cPanel problem now but can post an update if I get a satisfactory explaination from my hosting company if anyone is interested?
     
Loading...

Share This Page