horizon on vip__public becames unavailable after warm reboot of primary controller

[Ekaterina Khomyakova]

Fuel 9.2 #771, dvs 3.1 #229

Steps to reproduce:
  1. Prepare master (on 10.109.5.2 (as the second env)) with plugin, configure vCenter
  2. Add nodes with the following roles:
       * Controller
       * Controller
       * Controller
       * Compute + Cinder
       * Cinder-VMWare
  3. Deploy environment
  4. Try to access the Horizon by following the link on Dashboard

Actual result:
  * Fuel redirects to http://10.109.8.6/ and it returns 503 Service Unavailable
  * `openstack servers list` (and much others) doesn't work from console (but e.g. `nova list` works)
  * Sanity OSTF (and all others) fail

Expected results:
  * Fuel redirects to http://10.109.6.3/, where the Horizon is actually is
  * openstack works
  * OSTF pass

[Oleksiy Molchanov (omolchanov)]

Please provide diagnostic snapshot or env.

[Ekaterina Khomyakova (ekhomyakova)]

Env was provided.

[Igor Gajsin (igajsin)]

Looks like there is the problem whithin our virtual infrastructure. There was no connectivity between the haproxy netns and b-mgmt. Now I can't reproduce that bug and wait when it reproduces.

[Igor Gajsin (igajsin)]

Please forget that I told before. The reason of that bug is absence of the route* for the managemnt network in the netns haproxy.

So I keep research of the wrong behaviour of the ocf script ns_IPaddr2

---------
* - like 10.109.1.0/24 via 240.0.0.1 dev hapr-ns metric 10000

[Igor Gajsin (igajsin)]

At least we have workaround: # crm resource restart clone_p_haproxy

[Maria Zlatkova (mzlatkova)]

MOS 9.2 release note: https://review.fuel-infra.org/#/c/30272.