yahoo-eng-team team mailing list archive

[Miguel Lavalle <miguel@xxxxxxxxxxxx>]

I lean towards Neutron not having to handle infra-structure, especially
since the split brain partition state should be a transient and can be
handled by the deployment admin. I am marking this bug as won't fix. If
you think more discussion is warranted please feel free to change the
satus

** Changed in: neutron
       Status: Incomplete => Opinion

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to neutron.
https://bugs.launchpad.net/bugs/1877248

Title:
  Rabbitmq cluster split brain causes l2 population related flows to be
  cleared when ovs-neutron-agent is restarted

Status in neutron:
  Opinion

Bug description:
   Pre-conditions: Rabbitmq split brain, and then restart ovs-neutron-
  agent

  results: In normal, when the ovs-neutron-agent restarts, the method
  add_fdb_entries will be called to refresh the l2 pop related flows.
  However, under my experimental conditions, the agent cannot  receive
  the rpc to call add_fdb_entries due to the Rabbitmq split brain, so
  the l2 pop related flows cannot be updated in time, so these old flows
  will be cleaned as stale due to the different cookie id. However,
  these flows are actually useful, and deleting them will affect the
  tenant traffic.

  Our temporary solution is to not restart the ovs-neutron-agent until
  the rabbitmq cluster is restored. Is there a better solution?

To manage notifications about this bug go to:
https://bugs.launchpad.net/neutron/+bug/1877248/+subscriptions