Zabbix 8 introduces native support for storing data history in ClickHouse, allowing users to offload measurement streams while maintaining standard data collection and trigger processing mechanisms. The system uses a HistoryProvider to map specific data types to either SQL or ClickHouse storage.
The integration supports uint, dbl, str, log, text, and json data types, but excludes binary values. Zabbix proxy does not use ClickHouse for local storage, instead forwarding collected values to the main server. Data is transmitted via HTTP/HTTPS using the RowBinary format for inserts and JSONCompactEachRow for queries.
A key limitation in the current implementation is that Zabbix does not calculate native trends for numeric data stored in ClickHouse. If long-term trends are required, numeric data should remain in the SQL database, while other types can be directed to ClickHouse.
Data retention in ClickHouse is managed via TTL settings rather than the Zabbix Housekeeper. Default scripts set a 31-day retention period with daily partitioning, though these can be customized. Physical data deletion occurs during background table merges, which should be considered for disk capacity planning.
