In docker/services/pacemaker/database/mysql.yaml we start a mysql_init_bundle container in order to initialize the mysql pacemaker bundle via puppet. If puppet fails, we don't fail the Ansible task, and thus the Overcloud doesn't fail. Eventually it will fail later when a service tries to actually use the database since one was never started.
this has already been fixed upstream: https://review.openstack.org/#/c/533062/ we need to backport all these for that to apply cleanly downstream: https://code.engineering.redhat.com/gerrit/#/q/topic:bug/1741345+(status:open+OR+status:merged)
QE testplan: - new deployment with 3 default HA Controller roles - new deployment with 3 standalone Database roles - update deployment with 3 default HA Controller roles from templates without patch to new templates with patch - update deployment with 3 standalone Database roles from templates without patch to new templates with patch - repeat all of the above tests with deployed-server - Negative test: force mysql_init_bundle to fail by adding a typo to the puppet manifest in docker/services/pacemaker/database/mysql.yaml for the mysql_init_bundle container. The overcloud deployment in Heat should then fail at that step. Some combination of the above could be used to reduce the testing matrix.
I've successfully tested deployments except split stack with 3 DBs and force failure that worked as expected. As per update DFG, updates are currently blocked. Tested on a puddle 2018-03-10.1
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHSA-2018:0602