[Pacemaker] Pacemaker fencing and DLM/cLVM

[Daniel Dehennin]

Christine Caulfield <ccaulfie at redhat.com> writes:

> It seems to me that fencing is failing for some reason, though I can't
> tell from the logs exactly why, so you might have to investgate your
> setup for IPMI to see just what is happening (I'm no IPMI expert,
> sorry).

Thanks for looking, but actually IPMI stonith is working, for all nodes
I tested:

    stonith_adm --reboot <node>

And it works.

> The logs files tell me this though:
>
> Nov 25 10:56:32 nebula3 dlm_controld[6465]: 1035 fence request
> 1084811079 pid 7358 nodedown time 1416909392 fence_all dlm_stonith
> Nov 25 10:56:32 nebula3 dlm_controld[6465]: 1035 fence result
> 1084811079 pid 7358 result 1 exit status
> Nov 25 10:56:32 nebula3 dlm_controld[6465]: 1035 fence status
> 1084811079 receive 1 from 1084811080 walltime 1416909392 local 1035
> Nov 25 10:56:32 nebula3 dlm_controld[6465]: 1035 fence request
> 1084811079 no actor
>
>
> Showing a status code '1' from dlm_stonith - the result should be 0 if
> fencing completed succesfully.

But 1084811080 is nebula3 and in its logs I see:

Nov 25 10:56:33 nebula3 stonith-ng[6232]:   notice: can_fence_host_with_device: Stonith-nebula2-IPMILAN can fence nebula2: static-list
[...]
Nov 25 10:56:34 nebula3 stonith-ng[6232]:   notice: log_operation: Operation 'reboot' [7359] (call 4 from crmd.5038) for host 'nebula2' with device 'Stonith-nebula2-IPMILAN' returned: 0 (OK)
Nov 25 10:56:34 nebula3 stonith-ng[6232]:    error: crm_abort: crm_glib_handler: Forked child 7376 to record non-fatal assert at logging.c:63 : Source ID 20 was not found when attempting to remove it
Nov 25 10:56:34 nebula3 stonith-ng[6232]:    error: crm_abort: crm_glib_handler: Forked child 7377 to record non-fatal assert at logging.c:63 : Source ID 21 was not found when attempting to remove it
Nov 25 10:56:34 nebula3 stonith-ng[6232]:   notice: remote_op_done: Operation reboot of nebula2 by nebula1 for crmd.5038 at nebula1.34bed18c: OK
Nov 25 10:56:34 nebula3 crmd[6236]:   notice: tengine_stonith_notify: Peer nebula2 was terminated (reboot) by nebula1 for nebula1: OK (ref=34bed18c-c395-4de2-b323-e00208cac6c7) by client crmd.5038
Nov 25 10:56:34 nebula3 crmd[6236]:   notice: crm_update_peer_state: tengine_stonith_notify: Node nebula2[0] - state is now lost (was (null))

Which means to me that stonith-ng manage to fence the node and notify
its success.

How the “returned: 0 (OK)” could became “receive 1”?

A logic issue somewhere between stonith-ng and dlm_controld?

Thanks.
-- 
Daniel Dehennin
Récupérer ma clef GPG: gpg --recv-keys 0xCC1E9E5B7A6FE2DF
Fingerprint: 3E69 014E 5C23 50E8 9ED6  2AAD CC1E 9E5B 7A6F E2DF
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 342 bytes
Desc: not available
URL: <https://lists.clusterlabs.org/pipermail/pacemaker/attachments/20141125/298af990/attachment-0003.sig>