After a statistics update, a critical reporting query suddenly runs ten times slower. Query Store shows that the query switched to a new plan with much higher duration. The DBA needs to restore the earlier performance immediately while investigating the root cause. What should the DBA do?
Show answer and explanation
Answer: C
Query Store keeps the history of plans for each query, including the earlier plan that performed well. Forcing that plan tells the optimizer to use it again, which restores the earlier performance immediately and buys time for root-cause analysis, such as reviewing the statistics change. This is a targeted, reversible fix. Dropping indexes, changing parallelism or restoring a backup would be disruptive, and none of them is needed to recover a single regressed query.
Why the other options are wrong
- A. Dropping nonclustered indexes removes access paths that other queries may depend on. It is a destructive change that could cause more regressions and does not target the plan that changed.
- B. Changing max degree of parallelism affects how every query uses parallel threads across the database. It is a blunt instrument that may not fix this plan regression and can hurt other workloads.
- D. Restoring from backup would discard recent data changes and cause long downtime. A single regressed query can be corrected by forcing its earlier plan, so this response is disproportionate.