Keystone is unavailable after blocking br-mgmt interface on first controller

Bug #1493045 reported by Vasily Gorin
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
Fuel for OpenStack
Triaged
High
Fuel QA Team

Bug Description

https://product-ci.infra.mirantis.net/view/7.0_swarm/job/7.0.system_test.ubuntu.ha_neutron_destructive/lastCompletedBuild/testReport/

build 284
Disconnect controllers and check pacemaker status is correct

        Scenario:
              Preconditions:
                 1. Create cluster
                 2. Add 3 nodes with controller roles
                 3. Add 2 nodes with compute roles
                 4. Deploy the cluster

            1. Block traffic on br-mgmt of the first controller
            2. Check pacemaker status
            3. Wait until MySQL Galera is UP on some controller
            4. Run OSTF

Expected:
All OSTF tests passed except one of it.

Actual result:
 OSTF tests failed during connection to keystone.

Revision history for this message
Vasily Gorin (vgorin) wrote :
Vasily Gorin (vgorin)
description: updated
Revision history for this message
Ihor Kalnytskyi (ikalnytskyi) wrote :

All checks on this job is failed.

  1/ ERROR: Verify connection is present and
  2/ FAIL: Disconnect controllers and check pacemaker status is correct
  3/ FAIL: Check 3 in 1 rabbit failover

Obviously, that's about our deployment and not about OSTF. Move it to Fuel Library.

BTW, I found the following info in the logs -

On the controllers node-5.test.domain.local and node-3.test.domain.local, resource master_p_rabbitmq-server has different list of nodes where it is started.. Please refer to OpenStack logs for more details.

Changed in fuel:
assignee: Fuel Python Team (fuel-python) → Fuel Library Team (fuel-library)
Changed in fuel:
assignee: Fuel Library Team (fuel-library) → Sergii Golovatiuk (sgolovatiuk)
Revision history for this message
Vladimir Kuklin (vkuklin) wrote :

From what I see in the logs it seems that you are picking incorrect nodes for tests. E.g. for Check 3 in 1 rabbit you are comparing node-5 (which was stopped by corosync) and node-3 which is in a healthy partition. Please recheck that this is not the case for all other tests also.

Changed in fuel:
status: New → Incomplete
assignee: Sergii Golovatiuk (sgolovatiuk) → Fuel QA Team (fuel-qa)
Revision history for this message
Sergii Golovatiuk (sgolovatiuk) wrote :

If you manually do

iptables -I INPUT -i br-mgmt -j DROP && iptables -I OUTPUT -o br-mgmt -j DROP

on primary controller, then try to run OSFT tests from UI. All tests will fail

If you unblock br-mgmt by

iptables -D INPUT -i br-mgmt -j DROP && iptables -D OUTPUT -o br-mgmt -j DROP

wait a minute then run OSTF and it will work normally.

Changed in fuel:
status: Incomplete → Triaged
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.