Details
-
New Feature
-
Status: Resolved
-
Major
-
Resolution: Duplicate
-
None
-
None
-
None
-
None
Description
Currently we have two log retention strategies: one based on time and one based on log size. These work well for "event" type data--i.e. data that consists only of appends. However if the events model changes to an underlying keyed data set, a more convenient retention strategy would delete keys that had been overwritten rather than retaining whole segments.
The proposed implementation would be a background process that scanned log segments and recopied only keys that hadn't been overwritten. Some more details are in this wiki:
https://cwiki.apache.org/confluence/display/KAFKA/Keyed+Messages+Proposal
Attachments
Issue Links
- duplicates
-
KAFKA-631 Implement log compaction
- Resolved