profiling-statement-fingerprints

To identify slow SQL statements without DB Console, analyze historical statement fingerprints from crdb_internal.statement_statistics. This approach surfaces slow, resource-intensive, or error-prone queries by evaluating latency, CPU, contention, and admission waits across time buckets.

Detecting plan instability involves analyzing historical statement fingerprints to identify multiple plan_hash values for the same query over time. By examining crdb_internal.statement_statistics, you can pinpoint when a single SQL fingerprint generates varying execution plans across different time buckets.

Troubleshoot admission control waits and memory spills by analyzing historical statement fingerprints for admission_wait metrics and resource contention. The data from crdb_internal.statement_statistics provides specific workflows to debug memory or disk spill investigations and admission control bottlenecks.

You can analyze contention and CPU usage metrics with per-database and per-application context support. The analysis of historical statement fingerprints handles mixed-collection metrics, including aggregated and sampled data, while satisfying JSON field handling and access permissions.

Find optimization opportunities by ranking historical SQL fingerprints to isolate resource-intensive patterns. The analysis references index recommendations when present within the historical statement statistics, helping you target queries with high latency, CPU usage, or contention for optimization.