cuibo01 commented on PR #5678:
URL: https://github.com/apache/hudi/pull/5678#issuecomment-1137001976

   > > > Thanks for the contribution, can we explain a little what's the 
purpose of this patch ? Is is an improvement or bug fix ? Personally i dislike 
to introducing multiple threads based on that StreamReadOperator is already 
with distributed parallelism.
   > > 
   > > 
   > > MergeOnReadInputFormat#open takes a long time, causing the chk to time 
out
   > > 
https://github.com/apache/hudi/blob/f30b3aef3e5e12e9119d6b5294c94a0282719e00/hudi-flink-datasource/hudi-flink/src/main/java/org/apache/hudi/table/format/mor/MergeOnReadInputFormat.java#L655
   > 
   > Thanks, did you analyze what causes the most of time here ? I guess it is 
the Merge On Read table iterator initialization takes much time, because it 
reads all the records of one file group into a map (`BitCaskMap` by default), 
which takes time(for precombining logic here).
   > 
   > Is the time out caused by too many splits queued up there or due to some 
file split is huge ?
   
![image](https://user-images.githubusercontent.com/36881649/170227596-8a09d8de-7760-4966-b3c9-b1cf583ec8e6.png)
   
   i find that large files cause slow opening...For example, when a job is 
restarted using `start-commit=earliest`


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to