VIP is not assigned to a unit with Vivid or Wily

[Ryan Beisner]

For all systemd-based targets (vivid-kilo, wily-liberty) the VIP is not assigned to a unit in hacluster + keystone deployments. The keystone API and CLI both fail because the VIP is unreachable. The Juju units are in a ready state and there are no charm hook errors. This has been confirmed on the stable and the next charms.

For upstart-based targets (trusty-*), with the same topology and charm configuration, behavior is normal and the VIP is assigned to 1 keystone unit.

# Reproducer, stable charms:
http://bazaar.launchpad.net/~ost-maintainers/openstack-charm-testing/trunk/view/head:/bundles/dev/hacluster-keystone-default.yaml

# Reproducer, "next" charms:
http://bazaar.launchpad.net/~ost-maintainers/openstack-charm-testing/trunk/view/head:/bundles/dev/hacluster-keystone-next.yaml

# Example:
http://paste.ubuntu.com/14079149/

See also the attached logs.tar, containing /var/log and /etc for every unit in the deployment.

This likely impacts other hacluster-XYZ combinations for Vivid and Wily, though I've not yet validated that claim.

[David Ames (thedac)]

The root cause is crmsh in vivid/wily/xenial is not compatible with pacemaker [1]

A secondary issue is charmhelpers.core.host.service_running() is not systemd compatible.

We' track LP:1445616 and I'll take a look at the charmhelpers issue.

[1] https://bugs.launchpad.net/ubuntu/+source/crmsh/+bug/1445616

[James Page (james-page)]

I've bumped the crmsh version in xenial - looks like we should re-sync pacemaker and corosync from Debian unstable as well once pacemaker gets through the NEW queue.

[James Page (james-page)]

As this is actually a bug in crmsh, not the charm, marking this bug as invalid