Automatic release of DLPAR and CoD resources after failure

There are two different methods for automatically releasing resources.

The first method (normal method) for automatically releasing resources is done during a planned takeover on the active node. Depending on your cluster topology, the release process is either synchronous or asynchronous. The releases process for the normal method consists of releasing DLPAR, On/Off CoD, and EPCoD resources. On/Off CoD resources are released after the DLPAR resources are released. EPCoD resources are released before the DLPAR resources are released for both synchronous or asynchronous release processes. Thus, the EPCoD resources are quickly made available to another frame.

The release process for the normal method is started from the active node before the standby node starts its acquisition process.

If the release process is synchronous, the DLPAR resources that belonged to the active node are released before the allocation process on the standby node starts. The synchronous release is required when both LPARs (active and standby) are hosted on the same frame. If the LPARs are hosted on different frames, the release of the DLPAR resources can be asynchronous because the standby node does not need the DLPAR resources. The asynchronous release process speeds up the takeover process because it is not required for the release process to be finished to start the acquisition of resources by the standby node.

The second method for automatically releasing resources is performed during an unplanned takeover. For this method, PowerHA® SystemMirror® provide another way for releasing resources called automatic release after failure. In this scenario, the failing node cannot perform the release process because it failed, and the release process is not done when the standby node starts its takeover process. This process might prevent the standby node from acquiring resources that are held by the failing node. If any EPCoD resources were held by the failing frame, those resources can be applied to the standby node (under certain condition) because the standby node can release these resources for the failing frame. When the failing frame is restarted, any DLPAR and On/Off CoD resources that are previously owned by that frame are automatically released.

During PowerHA SystemMirror startup, when a node that previously failed joins the cluster (either manually or automatically) PowerHA SystemMirror analyzes all resource group policy options and generates a plan for resource groups to be acquired and for the resource groups to be released. The resources that are released during the automatic release process might be required by the node.