[8.0] Pacemaker may to destroy on one of controllers after restart corosync

[Vadim Rovachev]

Detailed bug description:
On swarm test:
https://github.com/openstack/fuel-qa/blob/stable/8.0/fuelweb_test/tests/tests_strength/test_failover_base.py#L1210-L1238
we kill and run corosync on controllers 500 times in a row. But sometime this test hase failed.
Fail moment in test logs:
http://paste.openstack.org/show/495777/
fail moment on node-1(node with fail)
http://paste.openstack.org/show/495778/

failed job:
https://patching-ci.infra.mirantis.net/job/8.0.system_test.ubuntu.ha_neutron_destructive/8/

All snapshot attached.

Reproducibility:
sometime
https://mirantis.testrail.com/index.php?/cases/results/271655

[Oleksiy Molchanov (omolchanov)]

So, in logs we can see that actually node-1 have returned to cluster, it is also visible on node-2 (the node where from we run 'pcs status nodes' in tests). pcs_status command logs in diagnostic snapshot show that node-1 was in cluster. But test logs indicate that for 20 seconds node-1 was offline, so test failed.

I was trying to reproduce it, but didn't manage. So I am marking this as Incomplete, until we have environment to revert.

[Dmitry Pyzhov (dpyzhov)]

Bug is incomplete status for a month. Changing status to invalid. Please reopen the issue if you have more data.