DNS-HA fails to create DNS entries when invalid maas-api key is replaced with valid key

[Jeff Hillman]

In a situation where the bundle has an incorrect maas_credentials configured, the cluster will come up and show that is is Ready, and it will allow the underlying service (glance/keystone/mysql/whatever) to come up in a clustered state.

however, due to the API key being wrong, the dns entries are not created in MAAS.

If either a juju config hacluster-<service> maas_credentials=<correct-key> is entered -OR- the bundle is fixed and re-deployed on top of an existing cloud, the hostnames are still not created in MAAS.

I have tried manually running config-changed, updated-status, and even stop then start hooks and the DNS entries are never created.

This requires the model to be destroyed and then the bundle re-deployed.

[Liam Young (gnuoy)]

I think a fix for this is going to be dependant on https://review.openstack.org/#/c/605844/

[Jeff Hillman (jhillman)]

That's unfortunate, considering that the referenced bug has been stale for 3 months...

[James Page (james-page)]

the underlying issue here is that its not possible to update an existing resource that's already been created in pacemaker; an attempt was made to support this but it proved buggy in the underlying tooling so was reverted prior to the subsequent charm release.

[James Page (james-page)]

Resource mutation is covered in the hacluster charm in the duped bug report.