RollbackExecutor pauses the cluster for every rollback execution


Description

RollbackExecutor pauses the cluster for every rollback execution. This is excessive as it is only necessary to pause the cluster for upgrade rollbacks. Pausing the cluster causes certain platform features, such as previewing/committing update sets, to be disabled.

Steps to Reproduce

  1. Run an ATF test that creates or updates a large quantity of data.
  2. At the end of the test, while the lengthy rollback is taking place, attempt to preview/commit an update set.
  3. Observe the system will not allow you to do so.

Workaround

This problem is under review. You can Subscribe to this article to receive notifications when more information will become available.

If you are certain the cluster is not paused for an upgrade or upgrade rollback, you can un-pause the cluster by running gs.resume(); from the Scripts - Background module.

Check sys_rollback_context for records with a state of "Reverting" to see what is being rolled back. To verify there is no upgrade in progress, check the Upgrade Monitor module and the worker threads on all nodes.

 


Related Problem: PRB1307843