We need to subscribe to port before_delete events, because
on after_delete events the port is already removed from the db
and we can't retrieve the binding:host_id information (which
is not passed in the event either).
This change also include a unit test against this bug, which also
happens to go through the ML2 RPC-to-registry codepaths so that
the effect of real registry events against our code are tested, thus
testing the validity of the expectation that the bagpipe driver code
has against ML2. This addresses bug #1493901.
One test is not passing, and thus commented out: see bug 1519258.
Reviewed: https:/ /review. openstack. org/249081 /git.openstack. org/cgit/ openstack/ networking- bgpvpn/ commit/ ?id=2aebe3da2d1 6578574ab8f8c2a 86aeafdd16c0ca
Committed: https:/
Submitter: Jenkins
Branch: master
commit 2aebe3da2d16578 574ab8f8c2a86ae afdd16c0ca
Author: Thomas Morin <email address hidden>
Date: Tue Nov 24 09:51:47 2015 +0100
bagpipe: port delete action on BEFORE_DELETE event
Details on the issue in bug #1517480.
We need to subscribe to port before_delete events, because
on after_delete events the port is already removed from the db
and we can't retrieve the binding:host_id information (which
is not passed in the event either).
This change also include a unit test against this bug, which also
happens to go through the ML2 RPC-to-registry codepaths so that
the effect of real registry events against our code are tested, thus
testing the validity of the expectation that the bagpipe driver code
has against ML2. This addresses bug #1493901.
One test is not passing, and thus commented out: see bug 1519258.
Closes-Bug: 1517480
Closes-Bug: 1493901
Change-Id: Icce89d98bcc442 4529dcbe661fed3 c1c8191418c