-
Type:
Improvement
-
Resolution: Unresolved
-
Priority:
Minor - P4
-
None
-
Affects Version/s: None
-
Component/s: Compaction
-
Storage Engines, Storage Engines - Persistence
-
SE Persistence backlog
-
1
Inside the struct __wt_background_compact, we have the following fields:
uint64_t max_file_idle_time; /* File compact idle time */ uint64_t max_file_skip_time; /* File compact skip time */
Their description are not meaningful and one cannot know how they are used without digging into the code. They should be more descriptive:
- We use the first one as a threshold in seconds before we remove the file from the background compaction list. A file is idle if it's not reassessed by the thread. This usually happens when the file is deleted.
- The second one is in seconds, too. This is the time we wait before reassessing a collection that could not be compacted during the previous run so we don't try over and over on something that does not give good compaction results.