Splitting and merging

Splitting and merging optimize the use of nodes without your intervention.

OneFS monitors every node in a cluster. If a node is unreachable over the internal network, OneFS separates the node from the cluster, an action referred to as splitting. When the cluster can reconnect to the node, OneFS adds the node back into the cluster, an action referred to as merging.

When a node is split from a cluster, it will continue to capture event information locally. You can connect to a split node with SSH and run the isi event events list command to view the local event log for the node. The local event log can help you troubleshoot the connection issue that resulted in the split. When the split node rejoins the cluster, local events gathered during the split are deleted. You can still view events generated by a split node in the node's event log file located at /var/log/isi_celog_events.log.

If a cluster splits during a write operation, OneFS might need to reallocate blocks for the file on the side with the quorum, which leads allocated blocks on the side without a quorum to become orphans. When the split nodes reconnect with the cluster, the OneFS Collect system job reclaims the orphaned blocks.

Meanwhile, as nodes split and merge with the cluster, the OneFS AutoBalance job redistributes data evenly among the nodes in the cluster, optimizing protection and conserving space.