[CRON] Errored: apache/flink#40626 (master - d231032)

[Flink CI <fl...@foxmail.com>]

|  [

apache

/

flink

](https://travis-ci.org/apache/flink)  
---  
  
![branch icon](https://s3.amazonaws.com/travis-email-
assets/branch.png)[master](https://github.com/apache/flink/tree/master)  
  
|

![build hasfailed](https://s3.amazonaws.com/travis-email-assets/status-
failed.png)

[Build #40626 errored](https://travis-ci.org/apache/flink/builds/596081692)

![arrow to build time](https://s3.amazonaws.com/travis-email-assets/failure-
arrow.png)

![clock icon](https://s3.amazonaws.com/travis-email-assets/time.png)35 hrs, 19
mins, 58 secs  
  
---  
  
![caoyingjie
avatar](https://secure.gravatar.com/avatar/236491d8d4c8068f53073b6549b97ecf.jpg)caoyingjie

[d231032 CHANGESET
->](https://github.com/apache/flink/compare/17fae0e0273c031118bdedbcdb8a03c401257690...d231032595cfb292d2ed1c42725493e765210fc5)  
  
[FLINK-14118][runtime] Reduce the unnecessary flushing when there is no data
available for flush. (#9706)  
  
The purpose of this commit is to reduce unnecessary flushing when there is no
data available for flush. More specifically, when there is exactly one
BufferConsumer in the buffer queue of subpartition and no new data will be
added for a while in the future (may because of just no input or the logic of
the operator is to collect some data for processing and will not emit records
immediately), the previous implementation will continuously notify data
available (may wake up the netty thread), which is unnecessary, and
eliminating these unnecessary flush can reduce CPU usage by 20% - 40% for some
jobs.  
  
Benchmark results before the fix:  
  
Benchmark Mode Cnt Score Error Units  
networkSkewedThroughput thrpt 30 18240.197 ?? 1892.419 ops/ms  
  
After the fix:  
  
networkSkewedThroughput thrpt 30 24532.313 ?? 1118.312 ops/m