[Pacemaker] Nodes OFFLINE with "not in our membership" messages

[laurent+pacemaker at u-picardie.fr]

Andrew Beekhof <andrew at beekhof.net> writes:

>> I'm also impacted by this issue. (running pcmk 1.1.7 and corosync 1.4.4)
>> there's a closed bug report here : http://bugs.clusterlabs.org/show_bug.cgi?id=5040
>> as far as i understand it's an issue with coroync.
>>
>> Pacemaker 1.1.8 is supposed to have workarounds for it, but I was
>> still able to reproduce the bug when running it with corosync 1.4.4
>
> Thats because there were/are bugs on the corosync side of the fence too.
>>From pacemaker's point of view: garbage in, garbage out.
>
> Has there not been a 1.4 release since?

I've just tried pcmk 1.1.7 and corosync 2.1.0 and I'm still able
to reproduce a similar behavior: fenced node staying offline when
trying to rejoin the cluster.

>>
>> It's ok with pcmk 1.1.8 and corosync 2.1.0. But I do have some issues
>> with 1.1.8 that make it unsuitable for production env yet.
>
> Oh?

new thread in a minute or so :)

--