Autoprovisioning

When you add a node to an Isilon cluster, OneFS attempts to assign the node to a node pool. This process is known as autoprovisioning, which helps OneFS to provide optimal performance, load balancing, and file system integrity across a cluster.

A node is not autoprovisioned to a node pool and made writable until at least three nodes of the same equivalence class are added to the cluster. If you add only two nodes of an equivalence class, no data is stored on these nodes until a third node of the same class is added.

Similarly, if a node goes down or is removed from the cluster so that fewer than three equivalence-class nodes remain, the node pool becomes under-provisioned. In this case, the two remaining nodes are still writable. However, if only one node of an equivalence class remains, this node is not writable, but remains readable.

Over time, as you add new Isilon nodes to your cluster, the new nodes will likely be different from the older nodes in certain ways. For example, the new nodes can be of a different generation, or have different drive configurations. Unless you add three new nodes of the same equivalence class each time you upgrade your cluster, the new nodes will not be autoprovisioned.

To work around those restrictions, OneFS enables you to create three types of compatibilities: node class, SSD capacity, and SSD count. With the appropriate compatibilities in place, new node types can be provisioned to existing node pools. You can add nodes one at a time to your cluster, and the new nodes can become fully functioning peers within existing node pools.

For example, suppose a cluster had a node pool made up of three S200 nodes, and you purchase an S210 node. In addition to being a different node generation, the S210 node could have a different number and capacity of SSDs. With the appropriate compatibilities, the new S210 node can be provisioned to the S200 node pool.