ClickHouse has a bunch of "system" tables, which expose various pieces of its own state and behavior. For example, we currently enable the query_log, which tracks all SQL queries with details about their resource usage. We should consider enabling the system.part_log table. This keeps historical information about how ClickHouse manages and merges data parts, such as compacting data after an insert or applying a mutation (e.g., dropping old data because of a TTL). The system.merges table is always there, but this only contains active merges.
ClickHouse has a bunch of "system" tables, which expose various pieces of its own state and behavior. For example, we currently enable the
query_log, which tracks all SQL queries with details about their resource usage. We should consider enabling thesystem.part_logtable. This keeps historical information about how ClickHouse manages and merges data parts, such as compacting data after an insert or applying a mutation (e.g., dropping old data because of a TTL). Thesystem.mergestable is always there, but this only contains active merges.