[Pacemaker] Pacemaker not failing over correctly (DRDB/Heartbeat/Pacemaker/Mysql) on Centos 5.5

[Lars Ellenberg]

On Fri, Jan 28, 2011 at 02:22:22PM -0500, Brian Cavanagh wrote:
> Hi, 
> 
> I am having this issue where it appears that everything is working
> correctly, but when I simulate failure the failover fails to work correctly.
> the Migrate command works fine, I can transfer the service, and the error I
> get when a node is put into standby or a server goes down is
> 
> Any help would be greatly appreciated

Someone in #linux-ha just pointed us to this thread,
so I thought it should not go unanswered.

You are using Pacemaker version

> Version: 1.0.10-da7075976b5ff0bee71074385f8fd02f296ec8a3

and your DRBD RA complains about

> Jan 28 12:20:37 mdb3 lrmd: [2778]: info: rsc:drbd_mysql:1:65: stop
> Jan 28 12:20:37 mdb3 drbd[9631]: ERROR: you really should enable notify when using this RA

And now you are complaining about: but I _do_ have notify enabled.

Well, sure you do.

But since "stop" is a special action, pacemaker decided to treat it's
environment a little bit too special, and now the sanity check of
the DRBD RA, which should have prevented you from starting it without
notify enabled now fails only on stop, as there the environment suddenly
was different than expected.

This was fixed since, I believe it was fixed in Pacemaker, (and should
be released with 1.1.5), but was more robustly coded in the DRBD RA as
well.

So you can just upgrade your DRBD (which should provide you with the
updated resource agent as well), or, if you prefer, just grab the
resource agent script itself as a drop-in replacement.

http://git.drbd.org/?p=drbd-8.3.git;a=blob_plain;f=scripts/drbd.ocf


-- 
: 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.