[Openstack] rdo havana to icehouse: instances stuck in 'resized or migrated'
jpichon at redhat.com
Mon May 19 16:15:01 UTC 2014
On 19/05/14 11:48, Julie Pichon wrote:
> On 16/05/14 22:45, Dimitri Maziuk wrote:
>> Hi all.
>> Upgrading centos 6/rdo from havana to icehouse: I've updated all
>> services on the controller as per the fine manual, then updated one of
>> the 3 compute nodes.
>> Now I'm trying to migrate the (shut off) instances from havana compute
>> node to the icehouse compute node and they get stuck in
>> "resize_migrated". On the first one I "reset-state --active", renamed
>> /var/lib/nova/instances/<uuid>_resize and was able to issue a 'hard
>> reboot' from the dashboard.
>> Now I have the 2nd one stuck in the same state. I restarted compute with
>> debug on on the target (icehouse) compute node, there's nothing
>> interesting there so the problem is somewhere else. In fact, it looks
>> like all it is is something isn't popping up the "confirm resize" button
>> in the dashboard -- /var/lib/nova/instances is nfs-mounted so there's no
>> actual data resizing/migration necessary.
> Hi Dimitri,
> From a dashboard perspective, the "Confirm resize" button is only
> displayed when the instance status gets to the final "VERIFY_RESIZE"
> stage. Out of curiosity, does the confirm resize command works on the
> stuck instances from the CLI? ("nova resize-confirm <server_id>")
> I'm not familiar enough with Nova to be able to provide additional
> information on why the instances might be getting stuck, unfortunately,
> hopefully someone else can offer more insightful advice.
I had a chat with a Nova developer who pointed me to the following patch
at https://review.openstack.org/#/c/84755/ , recently merged in Havana
and included in the latest RDO Havana packages. Resize specifically is
one of the actions affected by this bug, you might want to check that
you're running the latest packages on your Havana node(s) and see if
this might help to resolve the problem?
>> I really don't want to go through the reset-state/rename dir dance for
>> every instance I have.
>> Any suggestions? Anyone done the havana-to-icehouse dance with 2 compute
>> Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
>> Post to : openstack at lists.openstack.org
>> Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
> Mailing list: http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
> Post to : openstack at lists.openstack.org
> Unsubscribe : http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack
More information about the Openstack