The default, preferred sequence followed by the Upgrade Manager to roll back a georedundant (three-server) cluster is as follows:
- Roll back the standby server.
- Roll back the spare server.
- Fail over to the standby server.
- Reapply the configuration to the cluster.
- Roll back the remaining server.
An optional advanced action lets you change this order if required. You can also roll back the second and third server in the cluster simultaneously.
The default, preferred sequence followed by the Upgrade Manager to roll back a two-server cluster is as follows:
- Roll back the standby server.
- Fail over to the standby server.
- Reapply the configuration to the cluster.
- Roll back the remaining server.
You can roll back up to four MPE or MRA clusters in parallel. (You must manually start the process for each cluster.)
To roll back a georedundant cluster:
- From the Upgrade section of the navigation pane, select Upgrade Manager.
The Upgrade Manager page opens.
- Select the cluster to be rolled back and click Start Rollback.
You are prompted: Are you sure that you want to perform this action? Initiate backout server_name (back).
- Click OK to continue.
The confirmation window closes and the Upgrade Manager performs pre-rollback checks and then rolls back the standby server, in the primary site. The
Upgrade Operation column displays the progress of the current action.
Note: The number of steps in any given action is determined outside of the Upgrade Manager and may vary from release to release.
When the standby server is rolled back, the
Upgrade Operation column displays the message
Initiate backout Completed Successfully at date_time and Alarm 70501 (CLUSTER_MIXED_VERSION) is asserted.
Tip: Review alarms associated with each action before proceeding with the next action.
- Select the cluster again and click Continue Rollback.
You are prompted: Are you sure that you want to perform this action? Initiate backout server_name (back).
- Click OK to continue.
If you stop a rollback at a point of operator intervention, you can continue it later, or resume upgrading it from there.
The confirmation window closes and the Upgrade Manager performs pre-rollback checks and then rolls back the spare server, in the secondary site. The Upgrade Operation column displays the progress of the current action. When the spare server is rolled back, the Upgrade Operation column displays the message Initiate backout Completed Successfully at date_time.
- Select the cluster again and click Continue Rollback.
You are prompted.: Are you sure that you want to perform this action? Failover to old version cluster_name (next).
- Click OK to continue.
The confirmation window closes and the cluster fails over to the standby server, which becomes the active server.
- Reapply the configuration to the cluster:
- (For an MPE cluster) From the Policy Server section of the navigation pane, select Configuration, select the cluster, and in the Policy Server Administration page, click Reapply Configuration.
- (For an MRA cluster) From the MRA section of the navigation pane, select Configuration, select the cluster, and in the MRA Administration page, click Reapply Configuration.
The configuration information is synchronized.
- From the Upgrade section of the navigation pane, select Upgrade Manager.
The Upgrade Manager page opens.
- Select the cluster again and click Continue Rollback.
You are prompted: Are you sure that you want to perform this action? Initiate backout server_name (back)
- Click OK to continue.
The confirmation window closes and the Upgrade Manager performs pre-upgrade checks and then rolls back the remaining server, in the primary site. The Upgrade Operation column displays the progress of the current action.
When the remaining server is rolled back, the Upgrade Operation column displays the message Initiate backout Completed Successfully at date_time and Alarm 70501 is cleared.
The cluster is rolled back to the previous release. For each server, the Running Release column displays the release to which you rolled the system back, and the Up to Date column displays N (no).
If the rollback fails, a diagnostic message describes the problem. Try the rollback again; if it fails again, contact My Oracle Support.
Note: If the rollback fails and also leaves the server in an unrecoverable state (designated Zombie in the Upgrade Operation column, plus alarm 70508), the Upgrade Manager cannot resolve the issue. Contact My Oracle Support immediately.