[HA fencing plugin] Parallel deployment could cause not all of the STONITH primitives created
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Fuel for OpenStack |
Won't Fix
|
Medium
|
Bogdan Dobrelya | ||
6.0.x |
Won't Fix
|
Medium
|
Unassigned |
Bug Description
Fuel deploys non-primary controller nodes in Openstack environment in the parallel.
So, the Fencing plugin is being configured by the puppet in parallel as well.
As a result, some of stonith__* primitives and their location constraints could be absent in the Corosync configuration.
There is a workaround:
Once deployment is done, apply puppet manifest for plugin manually at affected controller nodes consequently:
cd /etc/fuel/
puppet apply -vd --modulepath=
(Although, the former apply action could be safely done for *any* controller node, as it would change or break nothing - even if the controller node is not affected by this issue)
Note:
You can list the STONITH primitives and check their health by the commands:
pcs stonith
pcs stonith level verify
And the location constraints could be shown by the commands:
pcs constraint list
pcs constraint ref <stonith_
It is expected that every stonith_* primitive should have one "prohibit" and one "allow" location shown by the ref command.
Changed in fuel: | |
status: | New → Triaged |
importance: | Undecided → Medium |
assignee: | nobody → Bogdan Dobrelya (bogdando) |
milestone: | none → 6.0.1 |
description: | updated |
Changed in fuel: | |
milestone: | 6.0.1 → 6.1 |
no longer affects: | fuel/8.0.x |
Changed in fuel: | |
status: | Confirmed → Won't Fix |
Related fix proposed to branch: master /review. openstack. org/147816
Review: https:/