Cluster Upgrade

In a clustered system, a rolling upgrade can be performed, eliminating downtime while the upgrade is performed. This section assumes familiarity with the Oracle ZFS Storage Appliance clustering model: if you are not familiar with the clustering concepts and terminology, please first read Chapter 10, Cluster Configuration, in Oracle ZFS Storage Appliance Administration Guide . To describe the rolling upgrade procedure, this document refers to the two clustered storage controllers as A and B, where A is the controller that will be updated first, and B is the controller that will be updated second. A key best practice in rolling upgrades is that each controller should be upgraded at a time when it is not providing service to clients. The procedure described here meets this requirement. In addition, all general upgrade best practices described earlier also apply to rolling upgrades.

Performing the Cluster Upgrade

Important: Do not perform a takeover operation while an upgrade is in progress.

1. Use the CLI or BUI to upload the software update image to both storage controllers.
2. Select which controller to update first. If a controller has no storage pools, update that controller first. In the following steps, controller A is updated first so if any clients are using controller A's storage pool(s), they experience a takeover-induced availability delay first.
3. Log in to controller A and use the CLI maintenance system reboot command or the BUI power icon in the masthead and select the reboot option to reboot controller A. Controller B takes over resources from controller A.
4. Log in to controller A and use the CLI or BUI to apply the software update to controller A. At the end of the upgrade, controller A reboots again and runs the new software version.

   ---

   Note -  Do not perform an upgrade on a controller while it is providing service.

   ---

5. Log in to controller B and use the CLI maintenance system reboot command or the BUI power icon in the masthead and select the reboot option to reboot controller B. Controller A takes over all resources and provides service using the new software version.
6. Validate the new software version on controller A and ensure that all services work correctly on the client systems.
7. If a serious problem appears, roll back controller A. Controller A reboots and controller B takes over and runs the previous software version. When controller A recovers, it also runs the previous software version.
8. If no serious problem appears, log in to controller B and use the CLI or BUI to apply the software update to controller B. Controller B reboots and runs the new software version.
9. Verify that all firmware updates complete.

   ---

   Note -  Controller firmware updates cannot proceed if the controllers are running different system software versions.

   ---

10. To restore normal operation and return resources to their respective assigned controllers, log in to controller A and use the CLI or BUI to perform a failback on controller A. For information on the failback operation, see Cluster Takeover and Failback in Oracle ZFS Storage Appliance Administration Guide .

Cluster States During Upgrade

The following table describes the state of the cluster after each step of the previous procedure.

Accessing the BUI or logging into the CLI while the controllers are running different software versions generates a warning that your configuration changes will not be propagated. You can configure the appliance to generate alerts when the cluster controllers are running different software versions (events "Cluster rejoin mismatch" and "Cluster rejoin mismatch on peer").

If you change the root password during an upgrade and then rollback the cluster, the nodes are not able to re-join after the rollback.