[SWARM][8.0] "Remove controllers" test detect outdated information in /etc/hosts file

[Vladimir Jigulin]

Reproduced on CI: https://patching-ci.infra.mirantis.net/job/8.0.acceptance.ubuntu.ha_scale_group_2/8/testReport/(root)/remove_controllers/remove_controllers/

Steps to reproduce:
1. Create cluster
2. Add 3 controller, 1 compute
3. Deploy the cluster
4. Remove 2 controllers
5. Deploy changes
6. Check /etc/hosts that removed nodes aren't present

Expected results:
Information about removed controllers must be deleted on other controllers in /etc/hosts file

Actual results:
/etc/hosts file on non removed controller contain information about removed controllers

Place in test:
https://github.com/openstack/fuel-qa/blob/stable/8.0/fuelweb_test/tests/tests_scale/test_scale_group_2.py#L174

Error message:
host node-2.test.domain.local is present in /etc/hosts

Reproducibility: 100%

[Alexey Stupnikov (astupnikov)]

I have checked another scenario: deployed one controller and then added 2 computes. As a result I have had correct /etc/hosts contents on all nodes. It looks like this bug is specific to node's delete process.

[Alexey Stupnikov (astupnikov)]

Confirmed for MOS8. Lowering this bug's severity to Medium, since IMO it doesn't break anything.

[Alexey Stupnikov (astupnikov)]

This bug was fixed during 9.0 release cycle (I can't reproduce it for clean MOS 9.0 env). We have discussed this issue on mos-maintenance team sync up and decided that I will try to find an easy fix. If it will take too much time to fix this issue, bug should be closed as won't fix.

[Alexey Stupnikov (astupnikov)]

Steps to reproduce:
1. Deploy env with 3 controllers and 2 computes.
2. Check /etc/hosts contents on any controller. There should be 2 records for every slave node.
3. Delete 2 controllers, deploy changes.
4. Check /etc/hosts contents.

Expected result: no records for deleted hosts.

Actual result: there are records for deleted hosts.

[Alexey Stupnikov (astupnikov)]

OK, this bug is a duplicate of bug #1513401. It is marked as won't fix, since the fix depends on feature's backport. Since this swarm issue is non-critical, we can't backport a feature to solve it. Closing as won't fix and marking as a duplicate of #1513401.