neutron-server fails to start on one or more controllers in HA scenario

[Giulio Fidente]

neutron-server occasionally fails to start on one or more controllers in the HA scenario, with pcs reporting the following:

  neutron-server_start_0 on overcloud-controller-0 'not running' (7): call=257, status=complete, exit-reason='none', last-rc-change='Wed Jun 17 12:55:58 2015', queued=2001ms, exec=3ms

and server.log printing:

  DBDuplicateEntry: (IntegrityError) (1062, "Duplicate entry 'datacentre-1' for key 'PRIMARY'") 'INSERT INTO ml2_vlan_allocations (physical_network, vlan_id, allocated) VALUES (%s, %s, %s)' (('datacentre', 1, 0), ('datacentre', 2, 0), ('datacentre', 3, 0), ('datacentre', 4, 0), ('datacentre', 5, 0), ('datacentre', 6, 0), ('datacentre', 7, 0), ('datacentre', 8, 0) ... displaying 10 of 1000 total bound parameter sets ... ('datacentre', 999, 0), ('datacentre', 1000, 0))

[Giulio Fidente (gfidente)]

neutron-server will try to populate the database with some data as soon as neutron-server is started; the error might be caused by a race due to multiple instances of neutron-server starting on different nodes

astapor seems to be doing a start/stop as workaround, see:
https://github.com/redhat-openstack/astapor/blob/master/puppet/modules/quickstack/manifests/pacemaker/neutron.pp#L87-L97

[Marios Andreou (marios-b)]

https://review.openstack.org/#/c/194610/ Add special handling of neutron-server service startup to fix race