Query Duration Tail Latency

warning

latencyUpdated Feb 6, 2026

High p95/p99 query duration indicates tail latency issues from inefficient queries, missing indexes, or resource contention, degrading user experience.

Sources

Essential Monitoring Queries - part 2 - SELECT Queries - ClickHouseclickhouse.com

Ignite ClickHouse Alerts: Detect Performance Issues Earlychistadata.com

How to Monitor ClickHouse Query Performance with OpenTelemetryoneuptime.com

Optimizing ClickHouse Performance: Diagnosing and Resolving ...www.atatus.com

Technologies:

ClickHouseSymptoms of this issue are visible in ClickHouse metrics and logs

clickhouse.query.duration_ms

clickhouse.query.memory_usage

clickhouse.query.read_rows

clickhouse.query.read_bytes

How to detect:

Track query_duration_ms from system.query_log at p95 and p99 percentiles. Alert when p99 exceeds acceptable thresholds (e.g., >1000ms for analytical queries).

Recommended action:

Identify slow queries via system.query_log, analyze execution patterns with send_logs_level='trace', optimize primary keys and projections, review ProfileEvents metrics to identify bottlenecks (I/O, CPU, memory), and consider query rewrites or materialized views.