diff options
| author | uncleGen <hustyugm@gmail.com> | 2017-01-18 10:55:31 -0800 |
|---|---|---|
| committer | Shixiong Zhu <shixiong@databricks.com> | 2017-01-18 10:55:31 -0800 |
| commit | a81e336f1eddc2c6245d807aae2c81ddc60eabf9 (patch) | |
| tree | a8033ee2351c5bcb16da85740cf54acc8e6adfac /mllib/src | |
| parent | 569e50680f97b1ed054337a39fe198769ef52d93 (diff) | |
| download | spark-a81e336f1eddc2c6245d807aae2c81ddc60eabf9.tar.gz spark-a81e336f1eddc2c6245d807aae2c81ddc60eabf9.tar.bz2 spark-a81e336f1eddc2c6245d807aae2c81ddc60eabf9.zip | |
[SPARK-19182][DSTREAM] Optimize the lock in StreamingJobProgressListener to not block UI when generating Streaming jobs
## What changes were proposed in this pull request?
When DStreamGraph is generating a job, it will hold a lock and block other APIs. Because StreamingJobProgressListener (numInactiveReceivers, streamName(streamId: Int), streamIds) needs to call DStreamGraph's methods to access some information, the UI may hang if generating a job is very slow (e.g., talking to the slow Kafka cluster to fetch metadata).
It's better to optimize the locks in DStreamGraph and StreamingJobProgressListener to make the UI not block by job generation.
## How was this patch tested?
existing ut
cc zsxwing
Author: uncleGen <hustyugm@gmail.com>
Closes #16601 from uncleGen/SPARK-19182.
Diffstat (limited to 'mllib/src')
0 files changed, 0 insertions, 0 deletions