yahoo-eng-team team mailing list archive

[Assaf Muller <amuller@xxxxxxxxxx>]

Public bug reported:

The main use case of L3 HA is bouncing back from a machine (That is
running a L3 agent) dying. In this case, with bp/report-ha-router-master
merged, any active routers on that node will remain active in the
Neutron DB (As the dead agent cannot update the server of anything). A
backup node will pick up the routers previously active on the dead node
and will update their status, resulting in the Neutron DB having the
router 'active' on two different nodes. This can mess up l2pop as HA
router interfaces will now be arbitrarily hosted on any of the 'active'
hosts.

The solution would be that when a L3 agent is marked as dead, to go
ahead and change the HA router states on that agent to from active to
standby, and also to update the router ports 'host' value to point to
the new active agent.

Note: This bug is at least partially coupled with
https://bugs.launchpad.net/neutron/+bug/1365476. Ideally we could solve
the two bugs in two separate patches with no dependencies.

** Affects: neutron
     Importance: High
     Assignee: Mike Kolesnik (mkolesni)
         Status: New


** Tags: l2-pop l3-ha

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to neutron.
https://bugs.launchpad.net/bugs/1461148

Title:
  Setting L3 agent status to "down" should update HA router states

Status in OpenStack Neutron (virtual network service):
  New

Bug description:
  The main use case of L3 HA is bouncing back from a machine (That is
  running a L3 agent) dying. In this case, with bp/report-ha-router-
  master merged, any active routers on that node will remain active in
  the Neutron DB (As the dead agent cannot update the server of
  anything). A backup node will pick up the routers previously active on
  the dead node and will update their status, resulting in the Neutron
  DB having the router 'active' on two different nodes. This can mess up
  l2pop as HA router interfaces will now be arbitrarily hosted on any of
  the 'active' hosts.

  The solution would be that when a L3 agent is marked as dead, to go
  ahead and change the HA router states on that agent to from active to
  standby, and also to update the router ports 'host' value to point to
  the new active agent.

  Note: This bug is at least partially coupled with
  https://bugs.launchpad.net/neutron/+bug/1365476. Ideally we could
  solve the two bugs in two separate patches with no dependencies.

To manage notifications about this bug go to:
https://bugs.launchpad.net/neutron/+bug/1461148/+subscriptions