RE: Solaris 9 processes that just won't die .. nearly there

From: Murray Robert-rmurra01 (robertmurray_at_motorola.com)
Date: 09/01/04

  • Next message: Jack Matthews: "FW: Solaris 2.6 box hung on boot" 
    To: "'Casper ***'" <casper@holland.sun.com>, Murray Robert-rmurra01 <robertmurray@motorola.com>
Date: Wed, 1 Sep 2004 22:43:02 +0100

    Hi All
    Thanks to you all for the replies

    Casper suggested using pstack..
    on reading the man page I discovered pwdx, prints the processes current working directory

    # pwdx 21320
    mvfs: ERROR: view=a20029_cr217973 vob=/usr/vob/omc_Osp1 - View storage directory or control files unavailable
    pwdx: cannot resolve cwd for 21320: I/O error
    #

    This has isolated the issue, to either a corrupt Clearcase view or Vob
    which can happen.

    The process is not dead yet, but we have something to work with now.

    Other suggestions were

    The preap utility has no effect
    # /bin/preap 21320
    preap: process not defunct: 21320
    #

    Checked the /etc/inittab file , theres no instructions to respawn the process

    Thanks again

    -----Original Message-----
    From: Casper *** [mailto:casper@holland.sun.com]
    Sent: 01 September 2004 15:18
    To: Murray Robert-rmurra01
    Subject: Re: Solaris 9 processes that just won't die

    >Hi All
    >
    >We're having serious issues with processes that will not die.
    >Our developers are running builds, on a group of 4 Solaris 9 machines.
    >
    ># more /etc/release
    > Solaris 9 4/03 s9s_u3wos_08 SPARC
    > Copyright 2003 Sun Microsystems, Inc. All Rights Reserved.
    > Use is subject to license terms.
    > Assembled 25 February 2003
    > Solaris 9 Maintenance Update 4 applied
    >#
    >Some of these processes start running at priority zero.
    >We try to kill them with kill -9 pid, but it doesn't work.
    >We have tried to renice the processes to a lower priority, but it doesn't work.

    Processes that cannot be killed are generally stuck in the kernel
    some place.

    Have you tried pstack?

    Have you tried using "mdb -k" and looked at the kernel
    thread list for the offending processes?

            mdb -k
    > $<threadlist ! more +/programname
            ... look at stack trace ...

    Casper
    _______________________________________________
    sunmanagers mailing list
    sunmanagers@sunmanagers.org
    http://www.sunmanagers.org/mailman/listinfo/sunmanagers

  • Next message: Jack Matthews: "FW: Solaris 2.6 box hung on boot"