Transparent Upgrade

A Transparent Upgrade is one in which services run without interruption throughout the cluster, during the upgrade. In a Transparent Upgrade, service software remains running without the benefit of oversight by the ClusterResourceManager[1] or LocalResourceManager[2].

Nodes in the cluster are upgraded, either one at a time or all at once, and then control of the services is returned to the ClusterResourceManager[1] either at the end or incrementally at the end of each node upgrade.

This is distinct from a RollingUpgrade[3] in which case all services remain at all times under the supervision of the ClusterResourceManager[1] and LocalResourceManager[2].

CRM transparent upgrade quick howto

The CRM clusters are very easy to upgrade transparently. A quick howto follows, but first a word of

WARNING: Always first test the new release with the existing configuration on a test cluster.

Now that you read and understood properly the previous note, here are the steps for a transparent upgrade:

1. Set all resources to the unmanaged mode:

    # crm_attribute -t crm_config -n is-managed-default -v false 

   Note that after the release 2.0.8 the attribute name changed from is_managed_default.

2. Check that all resources are unmanaged:

    # crm_mon -1

   If some are not, then remove the attribute from those resources:

    # crm_resource -r <rsc> --meta -d is_managed

3. Now you can stop heartbeat, upgrade, start.

   NB: In case the heartbeat does not want to stop, just kill all the processes.

4. Put resources back into the managed mode:

    # crm_attribute -t crm_config -n is-managed-default -v true

See Also

RollingUpgrade[3]

---

This information provided courtesy of the Linux-HA project at http://linux-ha.org/