High CPU usage on (Ubuntu 8.04 LTS Server)

Discussion in 'General' started by HenryTheEight, Jul 11, 2008.

  1. HenryTheEight

    HenryTheEight New Member

    Hi

    I just installed a fresh ISPconfig following the perfect setup on a esx server vm.
    I sometimes get an intermittent problem in which my CPU usage (2 vm cores used) would suddenly both maxed out and I lose access to any serving web pages. During
    this time I also can't ssh (or lose ssh connection if I was logged on), everything jams up. I can't even access the console. Everything seems to hang up, this seems to happen randomly, but once it starts the only remedy is to reboot and the CPU goes back to low i.e. < 2%.

    Is anybody facing this problem, I just upgraded my server to Intel Q9450 server. I didn't have this problem before on my older servers (AMD X2) with Centos. Using Latest version of Ispconfig.

    Advice and help needed. Complete newbie and lost as to what is the problem or how to solve this. Anyone else experiencing this?
     
  2. falko

    falko Super Moderator ISPConfig Developer

    Any errors in your logs?

    Did you run
    Code:
    top
    to find out which processes are eating up your resources?
     
  3. HenryTheEight

    HenryTheEight New Member

    Sorry to be a complete newbie/dummy,

    What logs should I be looking at and where can I get these logs
    from Ubuntu.

    My Esx server logs did highlight high cpu usage in which all VM cpus
    on Ubuntu were at 100% for more than half and hour before I reset/reboot it.

    Please advise on the which log files and where, so that I can give
    you more info/feedback.

    Thank you for your patience.

    BTW I can't run 'top' as VM console/screen just hangs, no network access to ssh but
    CPUs are all maxed.
     
    Last edited: Jul 12, 2008
  4. till

    till Super Moderator Staff Member ISPConfig Developer

    Please login with the user root on the shell and execute the command "top" as falko suggested. This programs listed on top are using the most computing power.
     
  5. HenryTheEight

    HenryTheEight New Member

    Just to reiterate and please forgive me.

    The problem is quite intermittent/random in which I am not sure what triggers the sudden max cpu usage from almost idle. So there is no sure way for me to repeat the trigger. Sometimes it is due to ftp transfer, sometimes in the middle of using ispconfig (not certain on what cause this).

    When cpu is idle all is ok i.e. networking, console, ssh, apache/php webpages, ispconfig all works fine. 'Top' command registers no highly active processes and I can run all linux command. All seem to work normally.

    When this so call cpu glitch occurs (i.e. CPU on ESX show max usage), all interaction with the server freezes. SSH disconnects, Console unresponsive i.e. can't issue commands ie. type etc. (not even top) to see what has suddenly triggered or what processes has launch to cause this.

    Is there a log file within Ubuntu which has a log of the process prior to this event? If so, any advice on location.

    Thanks in advance.

    BTW FYI I am not using any GUI e.g. Gnome on Ubuntu 'Server', I have heard and experienced that if you did have a GUI is screws up ISPConfig and may result in similarly high CPU usage. All install procedures of the Perfect setup followed to the letter. Setup is virtually fresh.
     
    Last edited: Jul 12, 2008
  6. till

    till Super Moderator Staff Member ISPConfig Developer

    You should try to switch the clamav installation to clamd on your server:

    http://www.howtoforge.com/forums/showthread.php?t=16204

    This happens only on Fedora and the load is not that high, just 1.0 and it can be fixed easily by removing the last entry in shells.
     
  7. sbrummer

    sbrummer New Member

    Did you find any solution?

    Hi,

    I'm experiencing the exact same problem. Having installed Ubuntu Server 8.04 from cd 5 days ago on a vmware esx client, i'm now seeing this high cpu load problem.

    Today I just left it running, while I was searching forums like this one. After 27 minutes the load went down to 50%, and now web access and ssh logins seems to hang indefinatly. When it was at 100% load both simply timed out.

    The ESX server is running on a quad core xeon, my client os has 2 cpu's assigned.
    The server could only have been serving web pages when it hanged. Nobody was logged in.

    I will now reboot and check if I have installed all updates. I will try to install vmware tools also, if it's possible on ubuntu, and try to run with only one cpu.

    If somebody reading this knows about the problem I would like to hear about it.

    Thanks in advance.
     
  8. falko

    falko Super Moderator ISPConfig Developer

    Not sure, but maybe another kernel helps?
     
  9. sbrummer

    sbrummer New Member

    Trying with vmware tools

    Hi Falco,

    Thanks for the reply.
    For now I have managed to get vmware tools installed, by following this guide:
    http://www.gorillapond.com/2006/07/31/install-vmware-tools-on-ubuntu/
    It has now been running stable for two days, but it's to early to say if the problem is gone.
    If it hangs again, I will try to change it to use only one cpu.
    I'll post here about my progress.
     
  10. sbrummer

    sbrummer New Member

    Installing vmware tools seems to be the solution

    Hi all,

    My server has now been running for 6 days without problems, so it seems the problem was solved when I installed vmware tools.
     
  11. sbrummer

    sbrummer New Member

    The high cpu load still happens occasionally

    Here is my experiences over the past 6 months:

    I have been running another ubuntu server under vmware esx at a hosting center for 6 months now. In the beginning it often "stalled" in the high cpu load state, but luckily for me I've got an online management console where I could hit the reset button.

    Trying to trace the error has not yielded any results yet. We have established two remote servers to trace the accessibility of the vmware guest server, and for a while it seemed like the "fixed interval" traffic those two servers generated would stabilize the guest server.
    But from time to time it still hangs, just not as frequent as in the beginning.

    My old setup is still running and it also stalls occasionally, so I can't say that installing the vmware tools as solved the problem.

    I have no idea of what can be the error, but it seems to only happen with the guests running ubuntu server.
    We have been running one guest with windows xp and another with "pbx in a flash" (centos) on the same esx server during the past 6 month, and they never stalled with the high cpu load problem.

    Searching the Internet now shows more people are experiencing the same problems with ubuntu server guests.
    So I really hope somebody can find a solution before next winter season starts, where my servers should be up 24/7 again.

    The stall problem still has the same symptons: Ping is possible, ssh and web simply stalls when connecting, I have never seen a timeout or got a response when the servers have stalled. Logs on the ubuntu guest server shows nothing strange at all.
     
  12. crypted

    crypted New Member

    Just for the hell of it, in SSH terminal watch TOP. See if at some point you see Spamassassin running high CPU usage rates. I have started having that issue after upgrading from 2.2.30 to 2.2.32 where my server entirely locks up.

    It is due to Spamassassin processes running indefinitely (hours) on large email files. There is a check in ISPCONFIG's user settings to keep this from happening but it stopped working.

    If you don't want to watch TOP to see what happens, go to ISPCONFIG's management, click on settings, EMAIL, and disable Spamfilter. Restart ISPCONFIG and whatever else, maybe a reboot, and just see if that clears it up.

    Then if that works you know that the problem is the same that I've been working on for two days now. And, we can diagnose better.
     
  13. sbrummer

    sbrummer New Member

    Top is not an option

    It's not possible to connect with either through ssh or esx admin console, when the error occurs. So I can't run top or any other on-system utility.

    It's can't be spamassassin, since it's not installed on those two servers (no mail server). I don't think it's a process at all, it has to be a kernel or driver fault to stall the system completely.

    I found a hint in the vmware KB I'm trying at the moment:

    Attach clocksource=acpi_pm to the boot string on smp systems should fix a potential hang.

    I have added that on both ubuntu servers and another virtual server running "pbx in a flash".
    I'll post back if it seems stable or not.
     

Share This Page