On 05/04/2016 01:21 PM, Michal Skrivanek wrote:
On 29 Apr 2016, at 18:49, Will Dennis <wdennis@nec-labs.com> wrote:
Answers inline below...
From: Michal Skrivanek [mailto:michal.skrivanek@redhat.com] what exactly did you do in the UI? Clicked on the node, and in the bottom pane, clicked on the "Upgrade" link showing there (the nodes also had an icon indicating that updates were available)
so..it was not in Maintenance when you run the update? You should avoid doing that as an update to any package may interfere with running guests. E.g. a qemu rpm update can (and likely will) simply kill all your VMs, I suppose similarly for Gluster before updating anything the volumes should be in some kind of maintenance mode as well No, the "Upgrade" link once clicked migrates any running VM off the target node onto another node, then sets the target node into Maintenance mode, and then performs the updates. ok, thanks for clarification, you got me scared:)
Once the updates are completed successfully, it re-activates the node and makes it available again. On the second and third nodes this coming out of Maintenance process experienced a problem with mounting the Gluster storage so it seems, and had the problems I'd indicated. it might be a question for gluster guys, it might be that the maintenance process is a bit different there Sahina, can you check/comment that?
Activation of hosts - after maintenance, w.r.t gluster - checks that glusterd is connected and returns Peer status as connected. The error seems to be in activating the gluster storage domain - hosted_storage. Did you see anything suspicious w.r.t this in vdsm logs? Regarding the Host being unresponsive with "Heartbeat exceeded" - we have a bug logged on this which is being investigated - https://bugzilla.redhat.com/show_bug.cgi?id=1331006. In this bug too, the host regains connectivity after ~ 2 mins time.
Thanks, michal
_______________________________________________ Users mailing list Users@ovirt.org http://lists.ovirt.org/mailman/listinfo/users