2-Node cluster graceful start/shutdown procedure. | Nutanix Community
Skip to main content

In a traditional Nutanix cluster at least 3 nodes are expected to form a cluster. There is an option however to form a single-node or a two-nodes cluster for ROBO (Remote Office/Branch Office) implementations or as a backup site.

 

The working of the two-node cluster is different from our usual 3 or more nodes cluster. Some examples are:

  1. We cannot expand a two-node cluster.

  2. Node removal is not supported.

  3. There is no cluster stop for 2 node cluster

 

So what is the graceful way to shut down and start a two-node cluster? 

How to shut down a 2-node cluster:

  1. Ensure that the cluster has data resiliency OK and can tolerate one node down from Prism.

  2. Stop user VMs - graceful shutdown.

  3. There is no cluster stop for 2 node clusters. Log in to a CVM using the nutanix account, and perform a graceful shut down of the CVM. 

  4. Wait for 5-10 mins, and then shut down the second CVM. 

  5. Shut down the hosts.

  6. NOTE: The above two commands used to shutdown the CVMs are different. Take a look at the document at the end for reference.

How to start a 2-node cluster:

  1. Power on the CVM from step 4 above.

  2. Wait until the single node cluster is up (around 20 mins).

  3. Power on the CVM from step 3 above.

  4. Wait for 20 mins. (Because 2 nodes cluster became 1 node cluster, then we need to wait for 20 min to change the status to 2 node cluster back. Usually, data resiliency is not restored for extent groups immediately after cluster shutdown or node down/off. and after (usually) 2 curator scans data resiliency is back to normal)

To take a look at the commands involved to shut down the two CVMs, refer https://portal.nutanix.com/page/documents/kbs/details/?targetId=kA00e000000LL66CAG

For more information on the two-nodes cluster:

Prism Web Console Guide v5.16: Two-Node Clusters

Confused or have questions regarding the implementation or requirements/dependencies? Check out Guide for single-node and two-node clusters for any clarifications.