[ovirt-users] HA not working when killing one node

Simone Tiraboschi stirabos at redhat.com
Tue Oct 11 09:37:30 UTC 2016

On Tue, Oct 11, 2016 at 11:28 AM, Felip Moll <lipixx at gmail.com> wrote:

> Hello,
> I have latest version of ovirt 4 installed on a Centos 7, 2 hypervisor
> nodes (rdkvm[1-2]) and 1 ovirt-engine node (rdhead1).
> GlusterFS is setup with 2 bricks on both rdkvm1 and rdkvm2.
> I have some vm's running on the infraestructure. I have HA enabled and
> ipmi is working.
> I manually poweroff force one node. Then I expect that VM's running on
> that node are started in the remaining node because of the HA setting.
> What happens is that ovirt starts the powered off node but not the
> vm's that remain shutted down.
> Is it the correct behaviour? How can I make that vm's are auto-started
> if they were running before hypervisor kill?
No, but it's a know issue:

The issue is that if you properly shut down the host, for some reason also
the VMs there will be properly shutdown (or at least we are detecting it
like that) and so the HA mechanism will not restart them thinking that it
was a voluntary shutdown triggered by the user.
It should properly work if you simply disconnect the network from the test
host: the host should be fenced and the HA VM restarted somewhere else.

Please see also:

> Regards,
> Felip M
> --
> Felip Moll Marquès
> Computer Science Engineer
> E-Mail - lipixx at gmail.com
> WebPage - http://lipix.ciutadella.es
> _______________________________________________
> Users mailing list
> Users at ovirt.org
> http://lists.ovirt.org/mailman/listinfo/users
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ovirt.org/pipermail/users/attachments/20161011/458ca3d0/attachment-0001.html>

More information about the Users mailing list