[Pacemaker] KVM VM's on internal DRBD volumes

Pawel Warowny warp at master.pl
Tue Jun 21 14:36:04 UTC 2011


Dnia Tue, 21 Jun 2011 15:22:14 +0200
Pieter Baele <pieter.baele at gmail.com> napisał(a):

Hi

> Hi, below part of the configuration + tidy (a lot is configured using
> DRBD MC)

From my very limited experience:

Kvm guests controlled by ocf:heartbeat:VirtualDomain are very fussy.
Sometimes they do not want to start or stop because of 
"connection reset by peer" errors from libvirtd and the like.
The same thing with migrations.
This problem occurs mainly while several hosts
starting/stopping/migrating at the same time.

So:
- I set batch-limit: 1, so every start/stop/migration action takes
  place one after another.
- stop-dealay parameter for start/stop/migration_from also helps (for
  now i set it to 10s and i'm thinking about bigger value)
- timeout for migration should be really big if your guest is
  heavily loaded

Nevertheless I suggest you create configuration with dlm, clvm,
gfs/ocfs2 for shared libvirt configuration files, cluster
aware lvm (hmm, ok this awareness is kinda limited ;]) and locking
mechanism which can save your life.

Best regards
-- 
Pawel Warowny

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 198 bytes
Desc: not available
URL: <https://lists.clusterlabs.org/pipermail/pacemaker/attachments/20110621/f78dd8ff/attachment-0004.sig>


More information about the Pacemaker mailing list