mirror of
https://github.com/facebook/rocksdb.git
synced 2024-11-27 11:43:49 +00:00
b9d65f5aa6
Summary: In level compaction, if the total size (even if compensated after taking account of the deletions) of a level hasn't exceeded the limit, but there are lots of deletion entries in some SST files of the level, these files should also be good candidates for compaction. Otherwise, queries for the deleted keys might be slow because they need to go over all the tombstones. This PR adds an option `deletion_ratio` to the factory of `CompactOnDeletionCollector` to configure it to trigger compaction when the ratio of tombstones >= `deletion_ratio`. Pull Request resolved: https://github.com/facebook/rocksdb/pull/6806 Test Plan: Added new unit test in `compact_on_deletion_collector_test.cc`. make compact_on_deletion_collector_test && ./compact_on_deletion_collector_test Reviewed By: ajkr Differential Revision: D21511981 Pulled By: cheng-chang fbshipit-source-id: 65a9d0150e8c9c00337787686475252e4535a3e1 |
||
---|---|---|
.. | ||
compact_on_deletion_collector.cc | ||
compact_on_deletion_collector.h | ||
compact_on_deletion_collector_test.cc |