Metrics

See live server counters, background-calculated resource metrics, and per-processor CPU profiling — three views of server-level performance.

What it does

The Metrics section exposes three views of server-level performance data.

Metrics queries system.metrics for live counters: active connections, threads, merges in flight, and hundreds of other real-time gauges. Values reset when the server restarts.

Async Metrics queries system.asynchronous_metrics, which are updated in the background on a schedule (not per-query). These include CPU load, memory usage, uptime, and OS-level stats. Useful for dashboards and alerting.

Profiler queries system.processors_profile_log to show CPU time broken down by query processor (pipeline step). Use it to find which stage of query execution is consuming the most CPU.

Pages

Using it

• Open /metrics for live gauges — active connections, threads, in-flight merges. These reset on server restart.
• Use /asynchronous-metrics for background-calculated resource stats (CPU load, memory, uptime, OS-level) suited to dashboards and alerting.
• Open /profiler to break CPU time down by query processor and find the pipeline stage consuming the most CPU.

Permissions & access

All three pages share the metrics feature id.

CHM_FEATURE_METRICS_ACCESS=authenticated

CHM_FEATURE_METRICS_ENABLED=false

[features.metrics]
enabled = true
access = "authenticated"

Configuration

No feature-specific configuration. The profiler page requires system.processors_profile_log to be populated, which depends on <processors_profile_log> being enabled in the ClickHouse server config.

Notes & limitations

• system.metrics values are live and reset on restart. They do not persist history. For trending, use system.asynchronous_metrics or system.metric_log (not yet surfaced in chmonitor).
• Async metrics are recalculated on a server-defined interval (typically every few seconds); values may lag slightly behind real-time.

Related