Skip to content

Reconciliation of state parameters fails sometimes #387

Description

@blaisemGH

Describe the bug

We observed 3 issues with the parameter state update:

  1. If the parameter update fails once, then it does not seem to recover and apply the correct parameters.
  2. It seems like sometimes the parameter update will restart the pod. The pod takes a couple minutes to start up, but the reconciliation starts immediately. Without the database online, the reconciliation fails, leading to the issue in 1.
  3. Sometimes the reconciler fails for a different reason that I could not identify

To Reproduce
a) Set the SGA max size to the size of the node. This will cause the update to fail. Request a larger node, so that the max sga value should be valid, and the reconciliation will still not take place. This was error 1.
b) After the failure in a), I updated the cpu count from 15 to 16. This caused the pod to restart, leading to error 2.

Following a) and b), I updated the cpu count to 14. This did not cause a pod restart, and the update was successful.

I then updated the cpu count back to 16. Nothing happened. This was error 3.

Expected behavior
The reconciliation of parameters should always update to the latest parameters and/or recover after a failure for valid parameters.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions