Hi, On Thu, Nov 25, 2010 at 10:51:56PM +0000, Dave Williams wrote: > > To follow up it appears that lrmd is aborting with the same error > > message after executing both "crm configure verify" AND "lrmadmin -C" > > > > Strace yields the following: > > > > lrmd: [336]: debug: on_receive_cmd: the IPC to client [pid:342] > > disconnected.\n > > \nGThread-ERROR **: Trying to recursively lock a mutex at 'unknown', > > previously locked at 'unknown'\naborting...\n" > > > > >From the client's perspective lrmadmin never sees this as the daemon is > > respawned (by crmd I presume) before you can manually execute the command > > again. > > > > "crm configure verify" calls lrmadmin in a tight loop as it iterates over > > cluster resources and so catches lrmd before it restarted and reports "not > > running". > > > > Looks like the g_thread_init() on its own isnt sufficient... > > > But this is what Senko's patch (2444.diff) fixes - so with that added it cures > the abort in both situations above. Now time to look at his potential ref > leak.
That patch doesn't cure the cause, just works around it. lrmd would just keep accumulating open IPC sockets. Thanks, Dejan > Dave > > > _______________________________________________________ > Linux-HA-Dev: [email protected] > http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev > Home Page: http://linux-ha.org/ _______________________________________________________ Linux-HA-Dev: [email protected] http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev Home Page: http://linux-ha.org/
