Open Stack

Hello,

After a manual failover from the master to the slave keepalived by
shutting down keepalived service on the master, then restarting it
again, the VRRP packets the master sends are "lost" or at least not
visible from the slave during couple of secondes / minutes, leading to
the IP being on the two instances during this lapse of time.

We are running OpenStack Juno, with two debian wheezy VMs as keepalived
instances inside the OS, on KVM. I diagnostic the problem doing the
manual failover and monitoring the IP and network with ``ip addr show``
and ``tcpdump -n -i any vrrp`` command inside the VMs.

I suspect the OpenStack network agent not routing correctly the VRRP
packets during this lapse of time.

Do you have any clue on how to fix the problem or diagnostic more the
problem?

Thank you,

-- aRkadeFR