#37 node kernel panic

closed
5
2004-08-27
2004-03-31
jsu2
No

The three nodes on my four node debian cluster has
locked up again. The init node seems to be up, but not
useable. Included is the bt output from kdb of the
nodes. Each node has similar output. The only
difference that I could tell was the process id.

Discussion

  • jsu2

    jsu2 - 2004-03-31

    kernel panic transcript from node 2

     
  • jsu2

    jsu2 - 2004-04-05

    Logged In: YES
    user_id=1010539

    Okay, I let the initnode up this weekend (without the other
    nodes). And, now I get out of memory errors. The system
    seems to be up, but all the processes seem to have been
    killed. So, the boxed can be pinged, but can't be accessed
    remotely.

    I have 2.5GB of RAM and like 1GB of swap. When I log into
    the console, the output of "free" shows only half the total
    available RAM is being used. So, I'm pretty sure the system
    isn't out of memory.

     
  • jsu2

    jsu2 - 2004-04-09

    Logged In: YES
    user_id=1010539

    Initnode lockup on 2004-04-09.

     
  • jsu2

    jsu2 - 2004-04-09

    Latest initnode lockup

     
  • jsu2

    jsu2 - 2004-04-26

    Logged In: YES
    user_id=1010539

    The cluster didn't crash per se this time, but it looks like
    the memory leak caused it to kill off a bunch of processes
    leaving the system useless. I turned off netdump this time
    which may be what kept it from actually crashing like
    before; I can't verify this, though. The screen shots are
    from the initnode. The top of the first picture is the
    output of "call show_free_areas".

    http://br.endernet.org/~akrowne/metacluster_dump-4/

     
  • Laura Ramirez

    Laura Ramirez - 2004-08-27

    Logged In: YES
    user_id=300036

    Aaron Krowne has confirmed that the cluster has not crashed
    recently.
    I am closing this bug. Johnny Healey, who is now in charge
    of the cluster can reopen it if the panics reoccur.

     
  • Laura Ramirez

    Laura Ramirez - 2004-08-27
    • status: open --> closed
     

Log in to post a comment.

Get latest updates about Open Source Projects, Conferences and News.

Sign up for the SourceForge newsletter:





No, thanks