Details
-
Improvement
-
Status: Done
-
Major
-
Resolution: Done
-
None
-
None
Description
Right now, we do a sync per record written in HDFS. This has performance penalties as it requires a Namenode contact. We get around this by letting the user define the sync policy in Flux, but that's less convenient and more sensible defaults could be created. We should use the batch size as a sensible default for the sync policy (i.e. sync every batch).
Attachments
Issue Links
- links to