This seems to be a weird bug.. The root cause is that one of the compute nodes has a misconfigured ceph client secret. So what happened is that when a VM is scheduled to that node, it would error out because the secret uuid can't be found. Nova-scheduler then tries to reschedule the VM to another host, but apparently the volume was deleted before the second host can start the VM.
-Simon On Sun, Dec 15, 2013 at 4:41 PM, John Griffith <[email protected]>wrote: > On Sun, Dec 15, 2013 at 12:02 PM, Xu (Simon) Chen <[email protected]> > wrote: > > It doesn't always happen, but only happens when I create a batch of 5 > VMs or > > more. A few of the VMs would fail to create and become ERROR state. > > > > By digging into the logs, it seems that the VM failed because the volume > > could not be attached, which was in turn due to the volume being deleted > for > > some reason. > > > > I am running an HA setup, but even if I shut every component to a single > > instance this would still happen. Any ideas? > > > > -Simon > > > > > > _______________________________________________ > > Mailing list: > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack > > Post to : [email protected] > > Unsubscribe : > http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack > > > Any chance of some cinder-volume logs for this? There should be > something in there to indicate the delete etc. >
_______________________________________________ Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack Post to : [email protected] Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
