Skip to main content

Remove nodes from an AFX storage system cluster

Contributors netapp-aherbin
PDFs

You might need to remove nodes from your AFX cluster for maintenance or hardware replacement. You can remove only one node at a time, and after removing a node, you must also remove its HA partner.

About this task

  • For AFX storage systems, all storage is contained within a single Storage Availability Zone (SAZ) that is accessible by every node in the cluster. As a result, when a node is removed, data does not need to be copied from the departing node. Instead, pointers to the volumes on the departing node are updated among the remaining nodes. This is referred to as a Zero Copy Volume Move (ZCVM) and happens instantaneously because no data is actually copied or moved.

    Note

    Before removing the node, manually handle any volumes that cannot be moved by ZCVM. These include volume clones, offline volumes, encrypted volumes whose original encryption key has been changed or rekeyed, and volumes restored using SnapMirror cloud. Audit staging volumes and root volumes are moved by ZCVM during the node unjoin process.

  • Successfully unjoining a node from the cluster does the following:

    • Marks the node ineligible and out of CLAM quorum.

    • Cleanly shuts the node down and moves ownership to a surviving node.

    • Cleans the RAID high-availability (HA) state on the node and deletes it from the cluster database.

Step 1: Prepare to remove nodes from an AFX cluster

Before you remove nodes from your AFX cluster, you should verify that the cluster is healthy and that there are no data LIFs on the node to be removed.

Steps

  1. Confirm that the cluster is healthy:

    cluster show

  2. Confirm that the node you want to remove is online and healthy:

    system node show -node <node> -health -epsilon

  3. View the LIFs on the node:

    network interface show -curr-node <node>

  4. Migrate data LIFs from the node:

    network interface migrate -vserver <SVM> -lif <LIF> -dest-node <destination_node> -dest-port <destination_port>

    Repeat this command for all data LIFs. Only cluster LIFs and node management LIFs should remain on the node.

  5. Verify that all data LIFs have been removed and only cluster LIFs and node management LIFs remain on the node:

    network interface show -curr-node <node>
What’s next?

After you have completed the preparation steps, you can remove the first node in the HA pair from the cluster and then remove its HA partner.

Step 2: Unjoin a single node from an AFX cluster

Remove a node from an AFX cluster only after you have completed the steps necessary to prepare the node for removal. After you remove a node from the cluster, you must then remove its HA partner.

Before you begin

It is recommended that you issue an AutoSupport message to notify NetApp technical support that you are removing a node from your cluster.

Steps

  1. Evacuate the node volumes:

    system node evacuate-volumes start -node <node_to_remove>

  2. Verify that the node volumes have been evacuated:

    system node evacuate-volumes show -node <node_to_remove>

    Wait until the pending count is 0 before proceeding. Do not move epsilon or eligibility until the volumes are fully evacuated.

    If necessary, you can use the system node evacuate-volumes abort -node <node> command to abort the evacuation process.

  3. Identify the node that holds epsilon:

    cluster show -epsilon true

    1. If the node that you want to remove holds epsilon, move epsilon to another node:

      cluster modify -node <surviving_node> -epsilon true

  4. Identify the current primary node:

    cluster ring show

    1. If the node you are removing is the current primary node, make it ineligible:

      cluster modify -node <node_to_remove> -eligibility false

  5. Show the home-node of remaining data LIFs:

    network interface show -home-node <node_to_remove>

  6. Change the home-node of the remaining data LIFs to another node:

    network interface modify -vserver <SVM> -lif <LIF> -home-node <surviving_node>

    Leave cluster LIFs and node management LIFs on the departing node.

  7. Disable takeover of the node to be removed:

    storage failover modify -node <node_to_remove> -enable-takeover-of false

  8. Verify that takeover is disabled for the node to be removed:

    storage failover show

    Takeover is not possible on the HA pair while failover is disabled.

  9. On a surviving, healthy node, set the privilege level to advanced:

    privilege set -level advanced

  10. Unjoin the node from the cluster:

    cluster remove-node -node <node_to_remove>

    Confirm the warnings. The job should end with Node remove succeeded.

What's next?

Repeat the previous steps to remove the HA partner from the cluster. You cannot remove additional nodes until the HA partner of the unjoined node is removed.

Step 3: Optionally, clean the removed hardware

If you are repurposing a node that you have removed from a cluster, you must erase the node configuration and initialize all disks before it can rejoin a cluster.

Steps

  1. During the boot process, when prompted, press Ctrl-C to display the boot menu.

  2. Select the boot menu option (4) Clean configuration and initialize all disks.

What's next?

You can use the cluster add-node command to reuse the HA pair in the same cluster.