Network connectivity breaks after primary controller's resetting

[Anna Babich]

Steps to reproduce:
1. Create two networks and two routers, attach net01 to router01 and net02 to router02
2. Launch vm1 in net01 and vm2 in net02 and associate floating IPs to them. Now, pings between these VMs by their floatings are successful
3. Ensure, that router01 is scheduled on a primary controller. Otherwise, move it from non-primary to primary one
4. Reset a primary controller and ensure that all routers are rescheduled from it to alive non-primary one.
5. Wait 5-10 minutes and start pings between vm1 and vm2 by their floatings again
Expected result: pings are successful
Actual result: pings don't go

The bug isn't always reproduced. The last time happened on ISO#301, cluster: neutron+vxlan, 3 controllers, 2 computes.

Neutron logs for controllers are attached.

[Eugene Nikanorov (enikanorov)]

Preliminary analysis shows that qr-port is created on both controllers and then it has created as UP on old controller that was rebooted and as DOWN (on vlan 4095) on the new one.

[Ann Taraday (akamyshnikova)]

I looked though logs and I can't identify the problem correctly with them.

I tried to reproduce this issue on the same env that it appeared first. I've reset 15 times primary controller - every time pings starts properly. I've done also destroy-start several times for primary controller and ping goes properly as well.

[Timur Nurlygayanov (tnurlygayanov)]

Kristina, could you please check this issue on new version of MOS?

Thank you!

[Kristina Berezovskaia (kkuznetsova)]

There are no such situations during acceptance testing. So move it to invalid