The upgrade sequence upgrades the Management Plane at the end. If required, you can continue to make configuration changes while the Management Plane upgrade is in progress.

Prerequisites

Verify that the NSX Edge cluster is upgraded successfully. See Upgrade NSX Edge Cluster.

NSX supports only an odd number of nodes in a cluster. A three-node cluster is the recommended configuration for fault tolerance. The upgrade process also includes taking a local backup of NSX Manager nodes that can be used to restore or rollback the system. Once all prechecks are complete, the system saves the configuration backup followed by local backup of all the nodes in the cluster. To check if the system saved the local backup, you can go to the root admin and check the following folder:
  • /image/backup/<unified_app_version>/cluster-node-backups - if the upgrade is from NSX 4.0.x and earlier versions.

  • /config_bak/backup/<unified_app_version>/cluster-node-backups - if the upgrade is from NSX 4.1.0 and later versions.

The rollback backup is also saved at /config_bak. If the local backup fails for any reason, the upgrade is stopped.

Procedure

  1. Backup the NSX Manager.
    See the NSX Administration Guide.
  2. Click Start to upgrade the Management plane.
  3. Accept the upgrade notification.

    You can safely ignore any upgrade related errors such as, HTTP service disruption that appears at this time. These errors appear because the NSX Manager node may be rebooting during the upgrade. You can continue to monitor the progress of the upgrade from the UI of any of the other NSX Manager nodes.

    If you are using a Virtual IP address, the UI remains accessible but you need to re-authenticate yourself after all the nodes have been upgraded.
  4. In case of upgrade errors, NSX may prompt you to roll back the upgrade. The rollback is performed on all the NSX Manager nodes:
    1. Run the following command from a root shell on all the NSX Manager nodes :

      /etc/init.d/corfu-server stop

    2. Run the following command as an admin user on all the NSX Manager nodes:

      node-rollback run-step step1_start_rollback

    3. Run the following command as an admin user on any one of the NSX Manager nodes:

      node-rollback run-step step2_restore_data

      If you encounter any errors, run the following commands:
      1. Execute corfu_tool_runner to delete the record from the registry table: 
        /opt/vmware/bin/corfu_tool_runner.py  -t RegistryTable -n CorfuSystem --port 9000 -o deleteRecord --keyToDelete='{"namespace": "CorfuSystem","tableName": "CompactionControlsTable"}'
      2. Run compaction verification:
        /opt/vmware/bin/corfu_compactor_upgrade_runner.py --runs 3 --lock false
      3. Stop corfu-server on all NSX Manager nodes.
      4. Start corfu-server on all NSX Manager nodes.
      5. Resume with the restore command as an admin user on any one of the NSX Manager nodes:

        node-rollback run-step step2_restore_data

    4. Run the following command as an admin user on all the NSX Manager nodes:

      node-rollback run-step step3_exit_rollback

    The rollback applies only to the Management Plane upgrade. Your NSX Manager nodes return to the version prior to starting the Management Plane upgrade.

What to do next