[Pacemaker] Master/Slave resource cannot start

[Andrew Beekhof]

On Mon, Aug 24, 2009 at 2:52 PM, Andrew Beekhof<andrew at beekhof.net> wrote:
> On Mon, Aug 24, 2009 at 2:33 PM, Diego
> Remolina<diego.remolina at physics.gatech.edu> wrote:
>> I was noticing this even before the 1.0.5 update right after I changed from
>> heartbeat to openais. I assume there may be some files in that folder from
>> back when I was using heartbeat which were causing the problem even with the
>> older pacemaker version.
>>
>> If I want to delete all files in /var/lib/pengine, which of the following
>> options should I use:
>
> I'd not do any.  It really shouldn't be relevant.
> Nothing that uses these directories has even started yet.
>

Can you (in openais.conf) disable logging to stderr and also logging to a file.
Make sure logging to syslog is turned on though.

I bet you wont have any further problems if you do this.

I looked again at the stack trace, and I see:

#1  0x0000003be088c74d in _L_lock_1685 () from /lib64/libc.so.6
#2  0x0000003be088c497 in __tz_convert () from /lib64/libc.so.6
#3  0x0000000000418a16 in _log_printf ()

Almost certainly this is caused by the following block:

	if (((logmode & LOG_MODE_FILE) || (logmode & LOG_MODE_STDERR)) &&
		(logmode & LOG_MODE_TIMESTAMP)) {
		gettimeofday (&tv, NULL);
		strftime (char_time, sizeof (char_time), "%b %e %k:%M:%S",
				  localtime (&tv.tv_sec));
...
	}

I've no idea how either of those two standard system functions is able
to cause a deadlock, but its the most likely explanation.
Perhaps file a bug with the libc guys.