Tempest test_trunk_subport_lifecycle may fail as reusing subports won't always work
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
networking-ovn |
Fix Released
|
Undecided
|
Daniel Alvarez |
Bug Description
When a subport becomes up in OVN, networking-ovn will copy the binding host from the parent to it. This will enable Neutron to bind it and therefore it will transition to 'ACTIVE'.
However, when this subport is removed from the trunk and added to a different one (which is bound to a different host), the subport won't ever transition to 'ACTIVE' as the original binding info wasn't cleaned up and OVN will never set the port as 'up'.
This is exactly what test_trunk_
- Creates server1 and server2 with one trunk each
- Assigns subports to trunk in server1
- Verifies that subports and trunks are all ACTIVE
- Removes subports from trunk in server1
- Assigns subports to trunk in server2
- Verifies that subports and trunks are all ACTIVE
This last verification won't succeed if server1 and server2 are bound to different hosts as the subport will retain the old binding information but ovn-controller won't bind it to the right chassis.
The solution is to clear the binding host info when the port goes down.
Changed in networking-ovn: | |
assignee: | nobody → Daniel Alvarez (dalvarezs) |
tags: | added: networking-ovn-proactive-backport-potential |
The reason why this is not failing in our gate is because the subport is moved to a different trunk but that trunk is in the same node as we're not running multinode jobs.