When HALO starts up, if a node is unreachable for any reason, currently the manager assumes the worst doesn't make any assumptions about the resources that might be running on it.
This is overly pessimistic however; often the node will be powered off and the manager can use that knowledge to know that it is safe to run the resources on its partner.
In fact, it is likely that HALO may have previously fenced the node, and knew that it was powered off previously -- it just "forgot" that knowledge when restarting.
When HALO starts up, it could check the power status of the remote servers and use that knowledge to better manage the resources.
When HALO starts up, if a node is unreachable for any reason, currently the manager assumes the worst doesn't make any assumptions about the resources that might be running on it.
This is overly pessimistic however; often the node will be powered off and the manager can use that knowledge to know that it is safe to run the resources on its partner.
In fact, it is likely that HALO may have previously fenced the node, and knew that it was powered off previously -- it just "forgot" that knowledge when restarting.
When HALO starts up, it could check the power status of the remote servers and use that knowledge to better manage the resources.