MySQL slow queries

What does this alert mean?

The rate of slow queries (exceeding long_query_time) is elevated.

Contributing factors

• Missing indexes on frequently queried columns
• Full table scans on large tables
• Lock contention from long-running transactions
• Inefficient query patterns after a schema or data volume change

Affected components

• Application services making database queries
• InnoDB buffer pool (cache thrashing from full scans)

Expected action

Identify the slow queries using the slow query log or EXPLAIN, then add indexes or rewrite the query.

Metric

Description & unit

Cumulative counter of queries exceeding long_query_time. mysqld_exporter wraps this in rate().

Unit: queries per second

Threshold rationale

Any sustained rate above 0 indicates queries exceeding the configured slow query threshold.

Normal behavior

Should be 0 or near-0 in a well-tuned database.

Screenshots

Silencing

Silence during batch jobs or ETL runs that are expected to run slow queries.

Expected frequency

Occasional — often follows schema changes, data growth, or missing index after a new feature deploy.

Severity guidance

Warning severity. Escalate if slow queries are causing connection pile-up or user-visible latency.

Impacted users

Users of services making the slow queries.

Things to check to determine severity

• Was there a recent schema change or new feature deploy?
• Has the data volume grown significantly?
• Is lock contention visible in SHOW PROCESSLIST?

Prometheus query

Dashboards

• MySQL Query Performance — Fill in: dashboard URL

Log queries

• MySQL slow query log — Fill in: log query URL

Steps

1. 1Enable slow query log if not active: SET GLOBAL slow_query_log = 1; SET GLOBAL long_query_time = 1;
2. 2Tail the slow query log: /var/log/mysql/slow.log
3. 3Digest with pt-query-digest: pt-query-digest /var/log/mysql/slow.log | head -100
4. 4For the top queries, run EXPLAIN: EXPLAIN SELECT ...
5. 5Look for type=ALL (full scan) in the EXPLAIN output — add an index on the filtered column

Useful commands

Guidance

Escalate if slow queries are causing thread pile-up or end-user latency alerts.

Slack channels

• #database-oncall — Fill in: channel URL

Alert rule

Tuning notes

If this fires too often for acceptable batch workloads, increase long_query_time rather than raising the alert threshold.