[ClusterLabs] Antw: [EXT] Re: temporary loss of quorum when member starts to rejoin

[Jehan-Guillaume de Rorthais]

On Wed, 8 Apr 2020 08:07:54 -0400
Sherrard Burton <sb-clusterlabs at allafrica.com> wrote:

> On 4/8/20 3:09 AM, Ulrich Windl wrote:
> >>>> Jehan-Guillaume de Rorthais <jgdr at dalibo.com> schrieb am 07.04.2020 um
> >>>> 23:02 in  
> > Nachricht
> > <20140_1586293370_5E8CEA7A_20140_452_1_20200407230230.1bc9b7b0 at firost>:  
> >> On Tue, 7 Apr 2020 14:13:35 -0400
> >> Sherrard Burton <sb-clusterlabs at allafrica.com> wrote:  
> > 
> > [...]  
> >> But the best protection is to disable pacemaker on boot so an admin can
> >> investigate the situation and join back the node safely.  
> > 
> > Actually before pacemaker we had a two-node cluster running without
> > problems. Maybe pacemaker should fix two-node mode. Requiring manual
> > intervention should be an absolute exception; instead the fenced node
> > should integrate to the existing cluster, or form a new cluster. Both
> > correctly, of course... 
> 
> i concur completely. requiring manual intervention mostly defeats the 
> purpose of running a cluster manager.

Well considering resources like a database or filesystem, an auto-failback of a
node with corrupted data might be dangerous.