[LU-2962] MGT umount locks up when taking down the file system.

• Status: Resolved
• Project: Lustre
• Affects Version/s: Lustre 2.4.0
• Type: Bug
• Priority: Critical
• Reporter: James A Simmons
• Assignee: Bruno Faccini (Inactive)
• Resolution: Cannot Reproduce
• Severity: 3

Comments

[Bruno Faccini (Inactive)]

Hello James,
I am afraid the vmcore is not enough and we'll also need the vmlinux (likely from kernel-debuginfo) and lustre-modules you run.

[James A Simmons]

I also pushed all the debug rpms into uploads/LU-2962/. These are the ones used for the test shot. Let me know if you need anything else.

[Bruno Faccini (Inactive)]

Thank's James, will get these new datas and let you know asap.

[Bruno Faccini (Inactive)]

James,
In the crash-dump you uploaded, there is no MGS mount nor Lustre modules loaded !!...
BTW it has been taken on Jan 16th, after only 224s of uptime:

      KERNEL: usr/lib/debug/lib/modules/2.6.32-279.14.1.el6.head.x86_64/vmlinux
    DUMPFILE: ./vmcore-mgs  [PARTIAL DUMP]
        CPUS: 8
        DATE: Wed Jan 16 14:21:18 2013
      UPTIME: 00:03:44
LOAD AVERAGE: 2.20, 1.44, 0.60
       TASKS: 285
    NODENAME: widow-mgs2
     RELEASE: 2.6.32-279.14.1.el6.head.x86_64
     VERSION: #1 SMP Wed Nov 21 13:02:32 EST 2012
     MACHINE: x86_64  (2327 Mhz)
      MEMORY: 16 GB
       PANIC: "[  224.469800] Oops: 0002 [#1] SMP " (check log for details)

I don't think this could be the crash-dump taken from the situation you described in this ticket ...
Can you double-check ??

[James A Simmons]

Sure as soon as admin comes in tomorrow I will get the proper core dump. Sorry about that.

[Bruno Faccini (Inactive)]

No problem James.

[James A Simmons]

I have bad news. The admin never got a crash dump We are going to have to reproduce it at a smaller scale.

[Bruno Faccini (Inactive)]

James, Do you agree that we can reduce ticket's priority to Major ?

[James A Simmons]

I figure the reason for this was that it only occurs are very large scale so it doesn't impact every one. Now it would be nice to see this fixed before the release I have a test shot this week end coming up so we can see if the problem still exist and if it does we can get a crash dump this time.

[James A Simmons]

For the last test shot we did not get a lock up so that is the good news. It did take 5 minutes to unmount the MGT but it is not a show stopper for us.

[Bruno Faccini (Inactive)]

James, what about this quite old ticket, can we close it as never reproduced ?

[James A Simmons]

Yes please close the ticket.

[Bruno Faccini (Inactive)]

Thanks for your feedback James.