Scale down fails in split-stack
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
tripleo |
Fix Released
|
High
|
Emilien Macchi |
Bug Description
In Train, we added support to execute Ansible tasks just before node(s) are deleted during a scale-down. e.g. RHSM unregister, Nova Compute service teardown, etc.
The issue is that these Ansible tasks only run on the nodes that are given as Nova UUID argument in "openstack overcloud node delete" command and later converted to Ironic hostnames so Ansible knows how to limit the playbooks to a certain group of hosts which are removed.
The lookup obviously fails in split stack since the nodes aren't in Ironic:
msg='IronicActi
The problem is a bit more than that, and also affects the new nova-less deployment which also doesn't use Nova UUID anymore.
A first step is to skip Ansible if Split-Stack or Nova-less is used in the deployment.
A second step will be to fix the lookup to return a correct hostname so Ansible can run on these nodes before the actual node removal.
Fix proposed to branch: master /review. opendev. org/701255
Review: https:/