View top wait events

Oracle

SELECT event, total_waits, time_waited, ROUND(time_waited/100,2) AS sec_waited
FROM v$system_event
WHERE wait_class <> 'Idle'
ORDER BY time_waited DESC
FETCH FIRST 20 ROWS ONLY;
-- For real-time analysis: v$active_session_history (ASH)

v$system_event gives cumulative wait events; v$active_session_history gives active session sampling for the last few minutes (requires the Diagnostics Pack license).

PostgreSQL

SELECT wait_event_type, wait_event, count(*)
FROM pg_stat_activity
WHERE wait_event IS NOT NULL AND state = 'active'
GROUP BY 1, 2
ORDER BY 3 DESC;

pg_stat_activity shows the current wait state; sampling it repeatedly (polling) reveals the most frequent wait events.

MySQL

SELECT event_name, count_star, sum_timer_wait,
       ROUND(sum_timer_wait/1000000000000, 2) AS sec_waited
FROM performance_schema.events_waits_summary_global_by_event_name
WHERE count_star > 0
ORDER BY sum_timer_wait DESC
LIMIT 20;

events_waits_summary_global_by_event_name gives total wait times per event (in picoseconds); wait instruments must be enabled.

SQL Server

SELECT wait_type, waiting_tasks_count, wait_time_ms,
       signal_wait_time_ms, wait_time_ms - signal_wait_time_ms AS resource_wait_ms
FROM sys.dm_os_wait_stats
WHERE wait_time_ms > 0
ORDER BY wait_time_ms DESC;

sys.dm_os_wait_stats shows cumulative wait types since instance startup; it can be reset with DBCC SQLPERF.

MongoDB

db.serverStatus().wiredTiger.concurrentTransactions
// For general waits/congestion:
// db.serverStatus().globalLock  and  db.serverStatus().wiredTiger.cache

WiredTiger has no classic wait events; exhaustion of concurrentTransactions (tickets), globalLock.currentQueue, and cache pressure indicate congestion.

Redis

CONFIG SET latency-monitor-threshold 100
LATENCY LATEST
LATENCY HISTORY command
# For the slowest commands: SLOWLOG GET 10

Once the latency monitor threshold is set, LATENCY LATEST/HISTORY reports the events causing the most latency.

Elasticsearch

GET _nodes/hot_threads

_nodes/hot_threads reveals bottlenecks by giving stack traces of the threads consuming the most CPU.