Running ClickHouse® in production requires visibility into cluster health. Here is what to monitor and why.
Key metrics to track
Every ClickHouse® deployment should monitor these metrics at a minimum:
System resources - CPU usage, memory consumption, disk space, and network throughput across all nodes. ClickHouse® is resource-intensive during merges and large queries.
Query performance - Track query duration, memory usage per query, and the number of concurrent queries. Slow queries often indicate missing indexes or suboptimal table schemas.
Merge operations - ClickHouse® continuously merges data parts in the background. Monitor merge progress, queue depth, and part counts per table. High part counts degrade query performance.
Replication lag - For replicated tables, monitor the replication queue size and lag between nodes. Replication issues can lead to inconsistent reads.
Alerting strategies
Not every metric needs an alert. Focus on actionable conditions:
- Disk usage above 80% on any node
- Replication queue growing for more than 10 minutes
- Any node unreachable for more than 60 seconds
- Query memory usage exceeding configured limits
SQL-based alerting, where the alert condition is a SQL query evaluated per-node, gives you the most flexibility and avoids generic threshold noise.
How CHOps helps
CHOps provides pre-built monitoring dashboards that cover all of these metrics immediately after connecting to your cluster. The query profiler with flame graphs helps identify slow queries, and DVR playback lets you rewind cluster history to investigate incidents after the fact.
Combined with SQL-based alerting and email notifications (Slack, Teams, PagerDuty in Pro), CHOps gives your team full observability without stitching together multiple tools.
