[Pacemaker] [Partially SOLVED] pacemaker/dlm problems

[Vladislav Bogdanov]

Ah, one (possibly important) addition.

That offending node is a vm on another cluster which experienced serious
problems that time due to bug in one of my RAs, and cluster host
carrying that vm was hardly shut down. So, it may be possible that
fencing request succeeded right before that host done. There was some
significant time before vm was started on another host. And it also
possible that vm failed again after start. Can't say anything more
precise right now, I needed to return all that to the alive state quickly.