No announcement yet.

Linux: server crashes hard, how to troubleshoot?

  • Filter
  • Time
  • Show
Clear All
new posts

    Linux: server crashes hard, how to troubleshoot?

    Hi all,

    this is my official request for help from all of you with experience setting up a Linux dedicated server.

    Whenever I start the Linux UT3 ded. server, a period of time passes after which the CPU is hogged, and the entire OS becomes unresponsive, and I have to restart the server *physically*. The period of time can be anything from 30secs to 10 minutes into a match.

    Steps taken:
    * During tests I run server with only critical processes running on the server
    * I have used the standard UT3 server configs for testing
    * I have set up hardware monitoring graphs (Munin) and checked spikes in I/O, temperatures, interrupts, etc.
    * I have checked logs for backtraces, weird IRQ events, whatever .. without luck.

    What are good steps to take in order to troubleshoot this?

    Server info:
    * Debian testing ("lenny")
    * Pentium 4 2.4GHz, 1GB RAM
    * Ded. 40/40Mbit connection
    * UT3 v1.2 (with Webadmin v0.9)

    root@isildur:~# uname -a
    Linux isildur 2.6.22-3-686 #1 SMP Mon Nov 12 08:32:57 UTC 2007 i686 pentium4 i386 GNU/Linux
    Anyone? What the h*ck is going on?

    What exactly is Debian testing? Perhaps a more mainstream distro would solve your problem?? That would likely be the fastest solution. Take a known working distro, any Fedora or others you find here that work without issues. I have installed the server on at least 6 machines, 2.4 P4 DFI board, 2.6 Celeron DFI board, 3.0 P4 intel board, 3700+ Machspeed, 3700+ ASUS A8v, 3700+ Gigabyte AGP, MSI 3700+ Neo4F, all with fedora 6, 7, and 8 with no issues.


      Originally posted by oldkawman1 View Post
      What exactly is Debian testing? Perhaps a more mainstream distro would solve your problem??
      I actually think you will find that Debian is quite mainstream. Not wanting to turn this into a distro-discussion, Debian is the basis of distros such as Knoppix, Mepis, **** Small Linux and Ubuntu. Debian testing is the version of Debian that has more recent (but admittedly less tested) versions of the kernel and software. It usually translates to the corresponding server/enterprise versions of, say, Mandriva and Fedora.

      Suffice to say that the kernel is not modified or specially compiled (nor is any other software on that machine), so I can't see how a different distro would solve the problem.

      I would really like to hear about any ideas for troubleshooting this problem, though. How would you go about finding the problem?