Unbatched Insert Pattern
warningconfigurationUpdated Feb 6, 2026
High-frequency small inserts create numerous small parts that degrade query performance and overwhelm the merge system, violating ClickHouse best practices.
Sources
Technologies:
How to detect:
Monitor InsertQuery rate from ClickHouseProfileEvents and correlate with parts creation rate from system.parts. Alert when insert frequency is high (>100/sec) but average rows per insert is low (<1000).
Recommended action:
Implement client-side batching to buffer inserts, use async_insert settings to enable server-side buffering, configure insert_quorum for consistency, or deploy a message queue (Kafka) to aggregate writes before ClickHouse ingestion.