Rejoining the cluster
When a node rejoins the cluster, the cluster managers running on the existing nodes initiate a node_up event to acknowledge that the returning node is up. When these nodes have completed processing their process_resources script, the new node then processes a node_up event so that it can resume providing cluster services.
This processing is necessary to ensure the proper balance of cluster resources. As long as the existing cluster managers first acknowledge a node rejoining the cluster, they can release any resource groups belonging to that node if necessary. Whether or not the resource groups are actually released in this situation depends on how the resource groups are configured for takeover (or dependencies). The new node can then start its operations.
Sequence of node_up events
The following list describes the sequence of node_up events:
- node_up
- This event occurs when a node joins or rejoins the cluster.
- process_resources
- This script calls the sub events needed for the node to acquire the service address (or shared address), gets all its owned (or shared) resources, and take the resources. This includes making disks available, varying on volume groups, mounting file systems, exporting file systems, mounting NFS file systems, and varying on concurrent access volume groups.
- process_resources_complete
- Each node runs this script when resources have been processed.
- node_up_complete
- This event occurs after the resources are processed and after the node_up event successfully completed. Depending on whether the node is local or remote, this event calls the start_server script to start application controllers on the local node, or allows the local node to mount an NFS file system only after the remote node is completely up.