[ClusterLabs] Query on HA

Andrei Borzenkov arvidjaar at gmail.com
Mon Aug 5 03:33:45 EDT 2019


There is no one-size-fits-all answer. You should enable and configure
stonith in pacemaker (which is disabled, otherwise described situation
would not happen). You may consider wait_for_all (or better two_node)
options in corosync that would prevent pacemaker to start unless both
nodes are up.

On Mon, Aug 5, 2019 at 9:52 AM Aloknath <mail2aloknath at gmail.com> wrote:
>
> Hi,
>
> I am beginner in Clusters and am facing one issue.
>
> I have a 2 node cluster designed for fail-over function. Pacemaker and corosync are cluster mangers used be build it.
>
> Scenrio :
>
> This happens  when both the nodes are powered off together and then again booted at same time.
>
> Issue:
>
> node2 booted little early and node1 is still in booting process . Here node1 resources are started in node2 (expected as node1 is still not online) then node2 brings up its own resources. During the same time node1 is online(fully booted) and starts to bring up its own resources .
>
> Now at the end NODE2 will have both node1 and node2 resources and  NODE1 also has its own resources . This is the issue am facing and this happens only when both nodes are booted at same time.
>
> But during pcs cluster standby / unstandby its works fine. It stops the resources in one node  and then only starts  it in other node. Even shutdown ,reboot all this scenario works fine if done on single node.
>
>
> Let me know if you need any more information .
>
> Regards
> Alok
>
>
> _______________________________________________
> Manage your subscription:
> https://lists.clusterlabs.org/mailman/listinfo/users
>
> ClusterLabs home: https://www.clusterlabs.org/


More information about the Users mailing list