Context
A two-node Oracle RAC primary with physical standby supported a production workload where availability and recovery remained the first priority.
Challenge, approach and acceptance.
Constraint
Connection failures and performance symptoms could have triggered speculative database changes even though the underlying cause might exist outside the database layer.
Approach
RAC workload evidence, listener behavior, connection patterns, system pressure and Data Guard state were reviewed together. A high-volume application retry pattern was isolated as a major contributor to listener-handler pressure.
Acceptance
The application-side behavior was corrected, connection stability recovered and database configuration was not changed without evidence that it was required.