[Pacemaker] Block stonith when drbd inconsistent

[Lars Ellenberg]

On Thu, Mar 07, 2013 at 11:09:03AM +1100, Andrew Beekhof wrote:
> If you put drbd into maintenance mode, we'd not perform any state
> changes (stop/stop/promote/demote) on it that could fail.
> That would likely do what you're after.
> 
> On Thu, Mar 7, 2013 at 4:59 AM, Jan Škoda <lefty at multihost.cz> wrote:
> > Dne 6.3.2013 06:38, Andrew Beekhof napsal(a):
> >> Nodes shouldn't be being fenced so often.  Do you know what is causing
> >> this to happen?
> > I know that this shouldn't happen frequently, but not having access to
> > uptodate data is certainly unwanted and there should be a way to prevent it.
> >
> > DRBD is quite prone to demote failures, especially when filesystem can
> > not be umounted for some reason. Blocked process for example can't be
> > killed and filesystems accessed by it can't be unmounted. This problem
> > is causing 90% of fencing for me.

So it is not DRBD failing to demote (go secondary),
but Filesystem failing to stop (umount),
that is causing your failures and fencing.

Why do you have "blocked processes"?
Maybe it is just a matter of chosing the right timeouts?


-- 
: 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.