Both cinder HA services not restarted when config changes
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
tripleo |
Fix Released
|
High
|
Alan Bishop |
Bug Description
When both cinder-backup and cinder-volume are deployed HA under pcmk, and a stack update causes a change in cinder's config, only the cinder-backup service is restarted. The cinder-volume service is not restarted.
The problem is due to the tripleo_ha_wrapper ansible role using a single marker file for detecting when the config changes. When processing the config changes for the cinder-backup service, the service is restarted and the marker file is updated to indicate the latest applied config. Then, when the cinder-volume service is processed, the marker file makes it appear as if the service is already running the latest config, and therefore the cinder-volume service doesn't get restarted. The tripleo_ha_wrapper ansible role needs to be told to use separate marker files for each cinder HA service.
Fix proposed to branch: master /review. opendev. org/c/openstack /tripleo- heat-templates/ +/849710
Review: https:/