[Pacemaker] help with DRBD/pacemaker

[Michael Schwartzkopff]

Am Dienstag, 14. Mai 2013, 00:52:32 schrieb Andreas Kurz:
> On 2013-05-13 11:54, Michael Schwartzkopff wrote:
> > Hi,
> > 
> > 
> > 
> > I experienced a strange phenomena when setting up DRBD 8.4.2 with
> > pacemaker 1.1.9.
> > 
> > 
> > 
> > I can set up a dual primary DRBD manually without any problems. Now shut
> > down one node, i.e. I demote the DRBD and the shut is "down".
> > 
> > 
> > 
> > When I start the cluster on the first node everything if fine. The CRM
> > detects that the DRBD is promoted. When I start pacemaker in the second
> > cluster node I see
> > 
> > 
> > 
> > - pacemaker starts DRBD nicely on the second node. It detects the
> > correct generation identifiers and the operations start and notify (2x)
> > show success.
> > 
> > 
> > 
> > Now the cluster wants to promote the second node and it generates a new
> > generation identified:
> > 
> > 2013-05-13T10:41:44.862939+02:00 xray kernel: block drbd0: role(
> > Secondary -> Primary )
> > 
> > 2013-05-13T10:41:44.862957+02:00 xray kernel: block drbd0: new current
> > UUID 35F2E1EB875F145F:2897B49E931464BC:6FE576F8519973CA:6FE476F8519973CB
> > 
> > 
> > 
> > Then DRBD starts the connection and DRBD detects the different primary
> > generation identifiers. The result is a split brain.
> > 
> > 
> > 
> > Any idea why DRBD 8.4.2 does not start the network connection when
> > starting the DRBD on the second node? Why does the cluster promote the
> > DRBD (and thus generating the new GI) WITHOUT the connection between the
> > DRBD?
> 
> IIRC we already had that discussion her on the list ... Pacemaker is too
> fast and is ready to promote faster then DRBD can connect (and nothing
> "forbids" it to do do so). But with correct resource-level fencing setup
> in DRBD that should work.

thanks. that solved the problem.

-- 
Dr. Michael Schwartzkopff
Guardinistr. 63
81375 München

Tel: (0163) 172 50 98
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.clusterlabs.org/pipermail/pacemaker/attachments/20130515/f52c1919/attachment-0003.html>