FeaturesMenu
On this page

Features

chmonitor is organized around the operational questions you answer while running a ClickHouse cluster.

Overview

Use the overview page first when you want a fast health check. It groups connections, queries, merges, replication, and system metrics into tabs so you can spot the noisy area before drilling into a table.

Query Monitoring

Use Running Queries for live work and Query pages for historical query log analysis.

  • Sort by elapsed time, memory, read rows, read bytes, and selected columns.
  • Open query text from table rows when you need the exact statement.
  • Use EXPLAIN tools to inspect plans and pipelines for a selected query.
  • Use the kill action only with a ClickHouse user that has KILL QUERY.

Tables and Storage

Use Tables, Top Usage Tables, and the explorer views when you need to understand storage growth.

  • Compare rows, bytes, parts, compression, and per-column sizes.
  • Use database and table filters before exporting large views.
  • Use table actions such as optimize only for tables where the monitoring user has the required ClickHouse grants.

Clusters and Replication

Cluster pages focus on distributed operational state:

  • distributed queues
  • replica lag and replication health
  • background merges and mutations
  • host-level CPU, memory, disk, and network signals

Data Explorer

The explorer is for ad-hoc read-only inspection. It can list databases, tables, columns, dependencies, and preview rows. Keep previews small when working with large production tables.

AI Agent

The agent can help turn natural language into ClickHouse questions, explain query behavior, and summarize findings. It still uses the same configured ClickHouse host and user, so its visibility is limited by the grants you give that user.

Required configuration:

LLM_API_KEY=your-provider-key
LLM_API_BASE=https://your-provider-compatible-endpoint/v1
LLM_MODEL=your-model

For deployments that expose the agent publicly, configure feature permissions so agent access requires authentication.

See the AI Agent guide for the full list of tools, skills, and example questions.

MCP Server

The MCP endpoint lets external AI tools query the monitor through /api/mcp. Production deployments require API key authentication. See the MCP Server reference before exposing this endpoint outside a trusted network.

Browser Connections

Browser connections let users store additional ClickHouse connection details in their browser and query through the app proxy. Use this for temporary inspection or personal workflows. For team dashboards, prefer server-side CLICKHOUSE_HOST, CLICKHOUSE_USER, and CLICKHOUSE_PASSWORD configuration.