HomeClusterLabs Projects

Test: scheduler: Update test outputs after clone instance shuffling fix

Description

Test: scheduler: Update test outputs after clone instance shuffling fix

The following tests are now correct:

  • clone-recover-no-shuffle-4
  • clone-recover-no-shuffle-5
  • clone-recover-no-shuffle-6
  • clone-recover-no-shuffle-7

Scores for several other tests are changed in ways (usually duplicates
from additional tentative assignments) that don't impact the resulting
transition.

One test (cancel-behind-moving-remote) technically breaks. Previously,
due to shuffling, ovn-dbs-bundle-1 moved to controller-0. Since
ovndb_servers:1 gets promoted on ovn-dbs-bundle-1, controller-0 held the
promoted instance of ovn-dbs-bundle.

Now, since instances correctly prefer their current nodes,
ovn-dbs-bundle-1 remains on controller-2. However, ovndb_servers:1 still
gets promoted on ovn-dbs-bundle-1, so controller-2 holds the promoted
instance of ovn-dbs-bundle.

ip-172.17.1.87 is colocated with ovn-dbs-bundle's promoted role and is
banned from controller-2. As a result, ip-172.17.1.87 is now stopped.

This test is believed to have worked properly in the past due only to
luck. At this point (see T672 and the bundle-promoted-*colocation-*
tests), it's well-established that colocations involving promotable
bundles don't work correctly.

Ref T489
Ref RHBZ#1931023

Signed-off-by: Reid Wahl <nrwahl@protonmail.com>

Details

Provenance
nrwahl2Authored on Jun 29 2023, 2:39 AM
Parents
rP512556ebd82c: Fix: libpacemaker: Don't shuffle clone instances unnecessarily
Branches
Unknown
Tags
Unknown
Tasks
Restricted Maniphest Task