Created attachment 480302 [details]
Preliminary patch; unit-tested but not reviewed.

Description of problem:

In some cases, it is desirable for resource agents to be able to time out or otherwise fail.  Typically, rgmanager places services where this occurs into the failed state.  This state requires administrative intervention in order for the service to exit this state.

Some resource-agents have a magic "self_fence" attribute, but this can fail for a variety of reasons - the most obvious one being use of queueing mechanisms when all paths fail to a given piece of storage.

In some cases, such as using HA-LVM with mirrored volume groups, there is a possibility for deadlock, causing self_fence to never be triggered.

The "hardrecovery" parameter could be used in conjunction with __enforce_timeouts in order to cause a failed -or- timed-out stop operation for a hung resource agent to cause the machine to reboot.

Version-Release number of selected component (if applicable): 2.0.115-68.el5