[
https://issues.apache.org/jira/browse/HUDI-1348?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Vu Ho updated HUDI-1348:
------------------------
Description:
Since DeltaStreamer makes heavily use of file listing, if the source contains a
lot of tiny files, this could quickly become a bottle neck. We need a way to
delete/archive files once processed by DeltaStreamer.
It seems like the best way to reliably clean up the source is after DeltaSync
commit the checkpoint successfully. We could add a new public method to Source
e.g. `postCommit()` and invoke it after each successful commit
Reference:
[https://spark.apache.org/docs/latest/structured-streaming-programming-guide.html#input-sources]
was:
Since DeltaStreamer makes heavily use of file listing, if the source contains a
lot of tiny files, this could quickly become a bottle neck. We need a way to
delete/archive files once processed by DeltaStreamer.
It seems like the best way to reliably remove/archive the processed files is
after DeltaSync commit the checkpoint successfully. We could add a new public
method to Source e.g. `postCommit()` and invoke it after each successful commit
Reference:
https://spark.apache.org/docs/latest/structured-streaming-programming-guide.html#input-sources
> Provide option to clean up DFS sources after each commit
> --------------------------------------------------------
>
> Key: HUDI-1348
> URL: https://issues.apache.org/jira/browse/HUDI-1348
> Project: Apache Hudi
> Issue Type: Improvement
> Components: DeltaStreamer, Utilities
> Reporter: Vu Ho
> Priority: Major
>
> Since DeltaStreamer makes heavily use of file listing, if the source contains
> a lot of tiny files, this could quickly become a bottle neck. We need a way
> to delete/archive files once processed by DeltaStreamer.
> It seems like the best way to reliably clean up the source is after DeltaSync
> commit the checkpoint successfully. We could add a new public method to
> Source e.g. `postCommit()` and invoke it after each successful commit
> Reference:
> [https://spark.apache.org/docs/latest/structured-streaming-programming-guide.html#input-sources]
>
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
