[SPARK-40932][CORE] Fix issue messages for allGather are overridden #38410
Conversation
This file contains hidden or 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
Contributor
Author
|
@jiangxb1987 could you help to review it? |
|
Can one of the admins verify this patch? |
Contributor
Author
|
@cloud-fan could you help to review this PR? |
cloud-fan
approved these changes
Oct 28, 2022
Contributor
|
thanks, merging to master/3.3! |
cloud-fan
pushed a commit
that referenced
this pull request
Oct 28, 2022
### What changes were proposed in this pull request?
The messages returned by allGather may be overridden by the following barrier APIs, eg,
``` scala
val messages: Array[String] = context.allGather("ABC")
context.barrier()
```
the `messages` may be like Array("", ""), but we're expecting Array("ABC", "ABC")
The root cause of this issue is the [messages got by allGather](https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/BarrierTaskContext.scala#L102) pointing to the [original message](https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/BarrierCoordinator.scala#L107) in the local mode. So when the following barrier APIs changed the messages, then the allGather message will be changed accordingly.
Finally, users can't get the correct result.
This PR fixed this issue by sending back the cloned messages.
### Why are the changes needed?
The bug mentioned in this description may block some external SPARK ML libraries which heavily depend on the spark barrier API to do some synchronization. If the barrier mechanism can't guarantee the correctness of the barrier APIs, it will be a disaster for external SPARK ML libraries.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
I added a unit test, with this PR, the unit test can pass
Closes #38410 from wbo4958/allgather-issue.
Authored-by: Bobby Wang <[email protected]>
Signed-off-by: Wenchen Fan <[email protected]>
(cherry picked from commit 0b892a5)
Signed-off-by: Wenchen Fan <[email protected]>
SandishKumarHN
pushed a commit
to SandishKumarHN/spark
that referenced
this pull request
Dec 12, 2022
### What changes were proposed in this pull request?
The messages returned by allGather may be overridden by the following barrier APIs, eg,
``` scala
val messages: Array[String] = context.allGather("ABC")
context.barrier()
```
the `messages` may be like Array("", ""), but we're expecting Array("ABC", "ABC")
The root cause of this issue is the [messages got by allGather](https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/BarrierTaskContext.scala#L102) pointing to the [original message](https://github.com/apache/spark/blob/master/core/src/main/scala/org/apache/spark/BarrierCoordinator.scala#L107) in the local mode. So when the following barrier APIs changed the messages, then the allGather message will be changed accordingly.
Finally, users can't get the correct result.
This PR fixed this issue by sending back the cloned messages.
### Why are the changes needed?
The bug mentioned in this description may block some external SPARK ML libraries which heavily depend on the spark barrier API to do some synchronization. If the barrier mechanism can't guarantee the correctness of the barrier APIs, it will be a disaster for external SPARK ML libraries.
### Does this PR introduce _any_ user-facing change?
No
### How was this patch tested?
I added a unit test, with this PR, the unit test can pass
Closes apache#38410 from wbo4958/allgather-issue.
Authored-by: Bobby Wang <[email protected]>
Signed-off-by: Wenchen Fan <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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.
What changes were proposed in this pull request?
The messages returned by allGather may be overridden by the following barrier APIs, eg,
the
messagesmay be like Array("", ""), but we're expecting Array("ABC", "ABC")The root cause of this issue is the messages got by allGather pointing to the original message in the local mode. So when the following barrier APIs changed the messages, then the allGather message will be changed accordingly.
Finally, users can't get the correct result.
This PR fixed this issue by sending back the cloned messages.
Why are the changes needed?
The bug mentioned in this description may block some external SPARK ML libraries which heavily depend on the spark barrier API to do some synchronization. If the barrier mechanism can't guarantee the correctness of the barrier APIs, it will be a disaster for external SPARK ML libraries.
Does this PR introduce any user-facing change?
No
How was this patch tested?
I added a unit test, with this PR, the unit test can pass