Background

Fencing in sharedroot environments should be done automatically. In some rare cases it might be usefull to fence a node by oneself. Here the possible ways to fence a node will be described.

Fencing directly

If you can login to the cluster and get a shell or only a filesystem that does not affect the node you are logged in you can easily fence the node by calling *fence_node *. After this the node is fenced and will be fenced again by the clusterinfrastructure and if everything is successful the journal will be replayed and the affected fs will unfreeze.

Fencing on a freezed node

If you cannot login to any node in the cluster you can still access fencing functionality via the fence_ackserver. This is a separate process running on any ATIX Sharedroot cluster. To access it you need a username and password (should be documented) and then you can login to the node via telnet on port 12242. Here you can also execute the fence_node command.