[ClusterLabs] Antw: Re: Antw: Re: A processor failed, forming new configuration very often and without reason

[Ulrich Windl]

>>> Philippe Carbonnier <philippe.carbonnier at vif.fr> schrieb am 17.04.2015 um 09:23
in Nachricht
<CAAfrj0xhU1rPFw-3PLWrHYyM3KQAhfw0hLBO7B4QxrrQSk7Hkw at mail.gmail.com>:
> Hello,
> 
> I use corosync for our customer, but I meet the problem "A processor joined
> or left the membership and a new membership was formed." without reason,
> I've asked the mailing list but even with the answer of Mr Beekhof the
> problem still occurs. Unfortunately no one can advice something on the
> mailing list so I'm asking for an enterprise who could help me to solve the
> issue, like RedHat tech support, but on corosync/pacemaker product. This is
> wha I call support.

Sorry, I can't help you, but I'd like to take the opportunity to ask for the following:

1) Ist there a network debugging tool at the TOTEM/corosync level? We see periodic "ring faulty" messages when the network is perfectly up.

2) There are many inter-dependencies of the RRP parameters. Ist there a kind of spreadsheet that will mark inconsistencies after you can entered your values? Maybe even if something like that exists, can it suggest new values, maybe if you want to add a node...

Recently we had the situation in a three-node cluster (A B C) where A and B stopped talking to each other (each felt as a DC) after the cluster stack was cleanly stopped on C (and C booted some DVD). Strange as can be...

Regards,
Ulrich