[Pacemaker] [Problem]Cib cannot update an attribute by 16 node constitution.

Andrew Beekhof andrew at beekhof.net
Mon Jun 14 06:21:32 UTC 2010


On Mon, Jun 14, 2010 at 4:46 AM,  <renayama19661014 at ybb.ne.jp> wrote:
> We tested 16 node constitution (15+1).
>
> We carried out the next procedure.
>
> Step1) Start 16 nodes.
> Step2) Send cib after a DC node was decided.
>
> An error occurs by the update of the attribute of pingd after Probe processing was over.
>
> ----------------------------------------------------------------------------------------------------------------------------------------
> Jun 14 10:58:03 hb0102 pingd: [2465]: info: ping_read: Retrying...
> Jun 14 10:58:13 hb0102 attrd: [2155]: WARN: attrd_cib_callback: Update 337 for default_ping_set=1600
> failed: Remote node did not respond
> Jun 14 10:58:13 hb0102 attrd: [2155]: WARN: attrd_cib_callback: Update 340 for default_ping_set=1600
> failed: Remote node did not respond
> Jun 14 10:58:13 hb0102 attrd: [2155]: WARN: attrd_cib_callback: Update 343 for default_ping_set=1600
> failed: Remote node did not respond
> Jun 14 10:58:13 hb0102 attrd: [2155]: WARN: attrd_cib_callback: Update 346 for default_ping_set=1600
> failed: Remote node did not respond
> Jun 14 10:58:13 hb0102 attrd: [2155]: WARN: attrd_cib_callback: Update 349 for default_ping_set=1600
> failed: Remote node did not respond
> ----------------------------------------------------------------------------------------------------------------------------------------
>
> In the middle of this error, I carried out a cibadmin(-Q optin) command, but time-out occurred.
> In addition, cib of the DC node seemed to move by the top command very busily.
>
>
> In addition, a communication error with cib occurs in the DC node, and crmd reboots.
>
> ----------------------------------------------------------------------------------------------------------------------------------------
> Jun 14 10:58:09 hb0101 attrd: [2278]: WARN: xmlfromIPC: No message received in the required interval
> (120s)
> Jun 14 10:58:09 hb0101 attrd: [2278]: info: attrd_perform_update: Sent update -41:
> default_ping_set=1600
> (snip)
> Jun 14 10:59:07 hb0101 crmd: [2280]: info: do_exit: [crmd] stopped (2)
> Jun 14 10:59:07 hb0101 corosync[2269]:   [pcmk  ] plugin.c:858 info: pcmk_ipc_exit: Client crmd
> (conn=0x106a2bf0, async-conn=0x106a2bf0) left
> Jun 14 10:59:08 hb0101 corosync[2269]:   [pcmk  ] plugin.c:481 ERROR: pcmk_wait_dispatch: Child
> process crmd exited (pid=2280, rc=2)
> Jun 14 10:59:08 hb0101 corosync[2269]:   [pcmk  ] plugin.c:498 notice: pcmk_wait_dispatch: Respawning
> failed child process: crmd
> Jun 14 10:59:08 hb0101 corosync[2269]:   [pcmk  ] utils.c:131 info: spawn_child: Forked child 2680 for
> process crmd
> Jun 14 10:59:08 hb0101 crmd: [2680]: info: Invoked: /usr/lib64/heartbeat/crmd
> Jun 14 10:59:08 hb0101 crmd: [2680]: info: main: CRM Hg Version:
> 9f04fa88cfd3da553e977cc79983d1c494c8b502
> Jun 14 10:59:08 hb0101 crmd: [2680]: info: crmd_init: Starting crmd
> Jun 14 10:59:08 hb0101 crmd: [2680]: info: G_main_add_SignalHandler: Added signal handler for signal
> 17
> ----------------------------------------------------------------------------------------------------------------------------------------
>
> There seems to be a problem in cib of the DC node somehow or other.
> We hope that an attribute change is completed in 16 nodes definitely.
>  * Is this phenomenon a limit of the current cib process?

More likely of the underlying messaging infrastructure, but I'll take a look.
Perhaps the default cib operation timeouts are too low for larger clusters.

>
> The log attached it to next Bugzilla.
>  * http://developerbugs.linux-foundation.org/show_bug.cgi?id=2443

Ok, I'll follow up there.

>
> Best Regards,
> Hideo Yamauchi.
>
>
>
> _______________________________________________
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker
>




More information about the Pacemaker mailing list