>>> Andrei Borzenkov <arvidj...@gmail.com> schrieb am 28.02.2020 um 06:00 in
Nachricht <60dcda9f-3f75-9fc4-8732-adc14ad3b...@gmail.com>:
[...]
>
> Well, this requires pacemaker supporting notion of "cluster wide
> shutdown" in the first place.
Yes, a "cluster shutdown" is very much desired, just as the "cluster startup"
(start all nodes at the same time conceptually). The latter is more than just a
ssh running the software on each node: The DC should be elected once a quorum
is up, then it should be waited for more nodes to join ("cluster formation
timeout"), and when either the timeout expired or all nodes are up, resources
should be placed (started).
Cluster shutdown should be orchestarted in a similar way: All resources
stopped, then nodes stop without unnecessary re-elections of the DC.
THAT would be a useful extension.
[...]
Regards,
Ulrich
_______________________________________________
Manage your subscription:
https://lists.clusterlabs.org/mailman/listinfo/users
ClusterLabs home: https://www.clusterlabs.org/
