[Pacemaker] DRBD Split Brain after each reboot

[Lars Ellenberg]

On Tue, Dec 22, 2009 at 09:26:44AM +0100, andschais at gmail.com wrote:
> Hi Michel,
> Yes, I have try with a simpler configuration, I follow this steps:
> 1) Master/Slave ocf:linbit:drbd RA + ocf:heartbeat:Filesystem RA-> shutdown
> -r now -> Ok, no split brain
> 2) ..+ ocf:heartbeat:IPaddr2 RA -> shutdown -r now -> Ok
> 3) ..+ heartbeat:drbdlinks RA -> shutdown -r now -> Ok
> 4) ..+ ocf:heartbeat:pgsql RA -> shutdown -r now -> Ok
> 5) ..+ ocf:custom:Asterisk RA -> shutdown -r now -> Ok
> 6) ..+ ocf:heartbeat:apache RA -> shutdown -r now -> Ok
> 7) ..+ lsb:postfix RA -> shutdown -r now -> Ok
> 8) ..+ lsb:dhcp3-server -> shutdown -r now -> Ok
> 9) ..+ lsb:lsb:atftpd -> shutdown -r now -> FAIL, Split brain
> At this point I get the first split brain, after a lot of google search I
> finally add a start-delay of ten seconds to Start and Promote operations for
> drbd RA. After that I reboot a couple times and everything works fine, no
> more split brain.
> 10) ..+ ocf:custom:JBoss RA -> shutdown -r now -> FAIL, Split brain
> With this resource enabled I always get a split brain after "normal" reboot.
> I tryed to
> increase start-delay time on both.start and promote operation to 40 seconds,
> that time is more than required to stop JBoss.
> If I remove start-delay , I can see in secondary logs:

No Sir.
increasing delays for drbd may be a workaround,
which works because our RA is retrying start and stop
for as long as it takes.
But your _problem_ is that the _other_ RAs apparently
pretend that something is stopped that actually is only "shutting down".

On Filesystem stop, or even on DRBD stop, you get a lot of log entries
about the device still being used, right?
well, fix that.

maybe you are just missing the correct order constraints?


-- 
: Lars Ellenberg
: LINBIT | Your Way to High Availability
: DRBD/HA support and consulting http://www.linbit.com

DRBD® and LINBIT® are registered trademarks of LINBIT, Austria.