Kirill Tkalenko created IGNITE-23291:
----------------------------------------
Summary: Improve the local compaction mechanism of Metastorage
Key: IGNITE-23291
URL: https://issues.apache.org/jira/browse/IGNITE-23291
Project: Ignite
Issue Type: Improvement
Reporter: Kirill Tkalenko
We need to improve local compaction of Metastorage, which is done in
*org.apache.ignite.internal.metastorage.impl.MetaStorageService#compact*.
* Perhaps it should be moved to another place.
* It should accept the revision to which keys should be compacted instead of
the timestamp.
* For each key, those versions whose revisions are less than or equal to the
revision of the compaction will be compacted (deleted). If the latest version
is equal to the compaction revision, it will not be compacted unless it's a
tombstone.
First example: The key "foo" had versions with revisions 1, 3 and 5. After
compaction to revision 3, the key will only have a version with revision 5.
Second example: The key "foo" had versions with revisions 1, 3 and 5. After
compaction to revision 10, the key will only have a version with revision 5.
Third example: The key "foo" had versions with revisions 1, 3 and 5
(tombstone). After compaction to revision 5 (or 10), all revisions will be
deleted.
*If a key has only one version, it will not be compacted (deleted).
For example: The key "foo" has only one version with revision 5. After
compacting to revision 5 or higher, this record will not be touched and will
remain with revision 5.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
