@Rocky Exactly what does the setting "lake_autovacuum_parallel_partitions" do on version 4.1.1?

In StarRocks version 4.1.1, the FE configuration setting

lake_autovacuum_parallel_partitions

determines the maximum number of partitions that can undergo the AutoVacuum process simultaneously in a shared-data (Cloud-native) cluster. Key Details * Default Value:

8

* Data Type: Int * Is Mutable: No (Changing this requires an FE restart) * Introduced In: v3.1.0 What it does In shared-data clusters, the AutoVacuum process acts as Garbage Collection (GC). After compactions are completed, older versions of data (rowsets/segments) are no longer needed. AutoVacuum identifies and deletes these obsolete files from the object storage to save space and reduce costs. This specific parameter controls the concurrency of this cleanup: * It limits how many partitions the Frontend (FE) will process for GC at any given time. * If you have a very large number of partitions frequently undergoing compaction, increasing this value (along with a restart) can speed up the reclamation of storage space. However, setting it too high may increase the load on the FE and the object storage API (e.g., S3/OSS delete requests). Related Settings If you are tuning AutoVacuum, you might also be interested in: *

lake_autovacuum_grace_period_minutes

: How long to keep historical data before it becomes eligible for vacuuming (default 30 mins). *

lake_autovacuum_stale_partition_threshold

: Prevents vacuuming on partitions that haven't been updated for a long time (default 12 hours). For more details, you can refer to the FE Configuration documentation. References • administration/management/FE_configuration.md • administration/management/compaction.md • sql-reference/System_variable.md • using_starrocks/async_mv/use_cases/data_lake_query_acceleration_with_materialized_views.md • faq/Others.md