force host should pass compute_filters
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
OpenStack Compute (nova) |
Confirmed
|
Low
|
Unassigned |
Bug Description
the compute service for opnstk1 is down
nova boot --availability-zone nova:opnstk1 --image a0c65331-
will make the instance stuck to
+------
| ID | Name | Status | Task State | Power State | Networks |
+------
| 7b0f8217-
+------
actually, we need to return the no valid host exception because host is down
Changed in nova: | |
assignee: | nobody → jichenjc (jichenjc) |
Changed in nova: | |
importance: | Undecided → Low |
status: | New → Confirmed |
Changed in nova: | |
assignee: | jichenjc (jichenjc) → Shashwat Srivastava (shashwat-srivastava) |
Changed in nova: | |
assignee: | Shashwat Srivastava (shashwat-srivastava) → nobody |
Changed in nova: | |
assignee: | jichenjc (jichenjc) → nobody |
status: | In Progress → Confirmed |
This seems to have the same cause as this bug: https:/ /bugs.launchpad .net/nova/ +bug/1427772 (the reason you see this is because ComputeFilter was not run).
However - you could still get this if you try to force an instance to a host before servicegroup API notices that the host has gone down. Since we use cast the result will be the same (the instance will be stuck with the task_state set to 'scheduling' forever.
I think that the only reasonable solution to this seeing how oslo.messaging works internally IMHO is to add a new task state (maybe 'dispatching to target node' or something to that effect) so I will not close it as a duplicate for now