[CELEBORN-1544][0.4] ShuffleWriter needs to call close finally to avoid memory leaks #2718
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…mory leaks
This PR aims to fix a possible memory leak in ShuffleWriter.
When we turn on `spark.speculation=true` or we kill the executing SQL, the task may be interrupted. At this time, `ShuffleWriter` may not call close.
At this time, `DataPusher#idleQueue` will occupy some memory capacity ( `celeborn.client.push.buffer.max.size` * `celeborn.client.push.queue.capacity` ) and the instance will not be released.
```java
Thread 537 (DataPusher-78931):
State: TIMED_WAITING
Blocked count: 0
Waited count: 16337
IsDaemon: true
Stack:
java.lang.Thread.sleep(Native Method)
org.apache.celeborn.client.write.DataPushQueue.takePushTasks(DataPushQueue.java:135)
org.apache.celeborn.client.write.DataPusher$1.run(DataPusher.java:122)
```
No
Production testing
<img width="547" alt="image" src="https://github.com/user-attachments/assets/d6f64257-144e-4139-96c6-518ca5f1bfd2">
<img width="479" alt="image" src="https://github.com/user-attachments/assets/e4ff62ec-5b9d-47a4-a36c-1d13bf378cbc">
Closes apache#2661 from cxzl25/CELEBORN-1544.
Authored-by: sychen <[email protected]>
Signed-off-by: zky.zhoukeyong <[email protected]>
…call abort to avoid memory leaks This PR aims to fix a possible memory leak in ShuffleWriter. Introduce a private abort method, which can be called to release memory when an exception occurs. apache#2661 Call the close method in the finally block, but the close method has `shuffleClient.mapperEnd`, which is dangerous for incomplete tasks, and the data may be inaccurate. No GA Closes apache#2663 from cxzl25/CELEBORN-1544-followup. Authored-by: sychen <[email protected]> Signed-off-by: zky.zhoukeyong <[email protected]>
cxzl25
approved these changes
Aug 30, 2024
|
thank you, merging to branch-0.4(v0.4.3). |
cfmcgrady
pushed a commit
that referenced
this pull request
Sep 2, 2024
…id memory leaks Backport CELEBORN-1544 (#2661 and #2663) to branch-0.4 ### What changes were proposed in this pull request? This PR aims to fix a possible memory leak in ShuffleWriter. ### Why are the changes needed? When we turn on `spark.speculation=true` or we kill the executing SQL, the task may be interrupted. At this time, `ShuffleWriter` may not call close. At this time, `DataPusher#idleQueue` will occupy some memory capacity ( `celeborn.client.push.buffer.max.size` * `celeborn.client.push.queue.capacity` ) and the instance will not be released. ```java Thread 537 (DataPusher-78931): State: TIMED_WAITING Blocked count: 0 Waited count: 16337 IsDaemon: true Stack: java.lang.Thread.sleep(Native Method) org.apache.celeborn.client.write.DataPushQueue.takePushTasks(DataPushQueue.java:135) org.apache.celeborn.client.write.DataPusher$1.run(DataPusher.java:122) ``` ### Does this PR introduce _any_ user-facing change? No ### How was this patch tested? Production testing #### Current <img width="547" alt="image" src="https://github.com/user-attachments/assets/d6f64257-144e-4139-96c6-518ca5f1bfd2"> #### PR <img width="479" alt="image" src="https://github.com/user-attachments/assets/e4ff62ec-5b9d-47a4-a36c-1d13bf378cbc"> Closes #2718 from pan3793/CELEBORN-1544-0.4. Authored-by: sychen <[email protected]> Signed-off-by: Fu Chen <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Backport CELEBORN-1544 (#2661 and #2663) to branch-0.4
What changes were proposed in this pull request?
This PR aims to fix a possible memory leak in ShuffleWriter.
Why are the changes needed?
When we turn on
spark.speculation=trueor we kill the executing SQL, the task may be interrupted. At this time,ShuffleWritermay not call close.At this time,
DataPusher#idleQueuewill occupy some memory capacity (celeborn.client.push.buffer.max.size*celeborn.client.push.queue.capacity) and the instance will not be released.Does this PR introduce any user-facing change?
No
How was this patch tested?
Production testing
Current
PR