PG-REC-WAITS-001 · Wait peaks during the recording

Severity	Confidence	Permission	Applies to
Medium	Hint	pg_monitor (pg_read_all_stats)	Recordings (codedb-lens record): PostgreSQL 14+, RDS, Aurora, Azure Flexible

What it checks

The activity collector: active sessions grouped by wait_event_type and wait_event (waits on the client left out). An episode is two or more samples in a row with 5 or more sessions on the same event (min_sessions).

Why it matters

For at least 30 seconds, several active sessions were waiting on the same event (locks, disk, WAL, internal locks). Each event points to a different bottleneck.

Next action

Match the peak with the event type: Lock = blocking between sessions, IO (DataFileRead) = reading from disk (missing indexes or memory), LWLock = contention inside the server (WAL, buffers), IO (WALWrite/WALSync) = WAL disk. Then look at the top queries of the same period.

How to verify

Record again during the same hours after the change. The peaks of that event must be lower or shorter.

Limitations

PostgreSQL shows only what each session waits on at the moment of the sample, not accumulated wait time, so short waits between samples are missed. A wait peak is a symptom, not a cause, so this is a hint.