Page MenuHomePhabricator

Applied puppet classes not appearing in horizon for integration-slave-docker-c2-m4-d40-1005.integration.eqiad.wmflabs
Closed, ResolvedPublic

Description

The role role::ci::slave::labs::docker is allowed on integration-slave-docker-c2-m4-d40-1005.integration.eqiad.wmflabs and appears on https://tools.wmflabs.org/openstack-browser/server/integration-slave-docker-c2-m4-d40-1005.integration.eqiad.wmflabs but does not appear as applied in horizon and thus can not be removed.

The same also happens for the integration-slave-docker-c2-m4-d40-1004.integration.eqiad.wmflabs instance that I created today, however this is a WIP instance and broken in other ways hence using the 1005 example.

Event Timeline

Re-adding the role via Horizon does not change anything. No error message for the add, but still not showing as applied on on any of the puppet tabs.

Interestingly, the view at https://tools.wmflabs.org/openstack-browser/puppetclass/role::ci::slave::labs::docker shows these roles being applied via wildcards:

integration-slave-docker-c2-m4-d40-1004.integration.eqiad.wmflab*
integration-slave-docker-c2-m4-d40-1005.integration.eqiad.wmflab*

Is this triggered by some limit on the instance name length?

I now remember (and just tested) that when I first created the 1005 instance I couldn't edit the hiera config.
That could also be due to the name length

That's exactly 64 characters there. Seems suspicious and plausible that it is about the max length.

Mentioned in SAL (#wikimedia-releng) [2017-10-17T22:08:26Z] <addshore> replaced integration-slave-docker-c2-m4-d40-1005 with integration-slave-docker-1005 T178409

So, I have removed both integration instances that were causing me issues and replaced them with instances with shorter names.

It seems the odd wildcard stuff still appears @ https://tools.wmflabs.org/openstack-browser/puppetclass/role::ci::slave::labs::docker and we might want to clean that up.

Andrew claimed this task.

I'm going to close this pending a re-appearance of the issue