[SPARK-35714][FOLLOW-UP][CORE] WorkerWatcher should run System.exit in a thread out of RpcEnv #35069
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
Member
Author
|
Thanks for proactively fixing this! |
Ngone51
added a commit
that referenced
this pull request
Jan 5, 2022
…n a thread out of RpcEnv ### What changes were proposed in this pull request? This PR proposes to let `WorkerWatcher` run `System.exit` in a separate thread instead of some thread of `RpcEnv`. ### Why are the changes needed? `System.exit` will trigger the shutdown hook to run `executor.stop`, which will result in the same deadlock issue with SPARK-14180. But note that since Spark upgrades to Hadoop 3 recently, each hook now will have a [timeout threshold](https://github.com/apache/hadoop/blob/d4794dd3b2ba365a9d95ad6aafcf43a1ea40f777/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/util/ShutdownHookManager.java#L205-L209) which forcibly interrupt the hook execution once reaches timeout. So, the deadlock issue doesn't really exist in the master branch. However, it's still critical for previous releases and is a wrong behavior that should be fixed. ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? Tested manually. Closes #35069 from Ngone51/fix-workerwatcher-exit. Authored-by: yi.wu <[email protected]> Signed-off-by: yi.wu <[email protected]> (cherry picked from commit 639d6f4) Signed-off-by: yi.wu <[email protected]>
Ngone51
added a commit
that referenced
this pull request
Jan 5, 2022
…n a thread out of RpcEnv ### What changes were proposed in this pull request? This PR proposes to let `WorkerWatcher` run `System.exit` in a separate thread instead of some thread of `RpcEnv`. ### Why are the changes needed? `System.exit` will trigger the shutdown hook to run `executor.stop`, which will result in the same deadlock issue with SPARK-14180. But note that since Spark upgrades to Hadoop 3 recently, each hook now will have a [timeout threshold](https://github.com/apache/hadoop/blob/d4794dd3b2ba365a9d95ad6aafcf43a1ea40f777/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/util/ShutdownHookManager.java#L205-L209) which forcibly interrupt the hook execution once reaches timeout. So, the deadlock issue doesn't really exist in the master branch. However, it's still critical for previous releases and is a wrong behavior that should be fixed. ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? Tested manually. Closes #35069 from Ngone51/fix-workerwatcher-exit. Authored-by: yi.wu <[email protected]> Signed-off-by: yi.wu <[email protected]> (cherry picked from commit 639d6f4) Signed-off-by: yi.wu <[email protected]>
Ngone51
added a commit
that referenced
this pull request
Jan 5, 2022
…n a thread out of RpcEnv ### What changes were proposed in this pull request? This PR proposes to let `WorkerWatcher` run `System.exit` in a separate thread instead of some thread of `RpcEnv`. ### Why are the changes needed? `System.exit` will trigger the shutdown hook to run `executor.stop`, which will result in the same deadlock issue with SPARK-14180. But note that since Spark upgrades to Hadoop 3 recently, each hook now will have a [timeout threshold](https://github.com/apache/hadoop/blob/d4794dd3b2ba365a9d95ad6aafcf43a1ea40f777/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/util/ShutdownHookManager.java#L205-L209) which forcibly interrupt the hook execution once reaches timeout. So, the deadlock issue doesn't really exist in the master branch. However, it's still critical for previous releases and is a wrong behavior that should be fixed. ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? Tested manually. Closes #35069 from Ngone51/fix-workerwatcher-exit. Authored-by: yi.wu <[email protected]> Signed-off-by: yi.wu <[email protected]> (cherry picked from commit 639d6f4) Signed-off-by: yi.wu <[email protected]>
Member
Author
|
Thanks all! Merged to Master/branch-3.2/branch-3.1/branch-3.0. |
Contributor
|
Late +1 |
fishcus
pushed a commit
to fishcus/spark
that referenced
this pull request
Jan 12, 2022
…n a thread out of RpcEnv ### What changes were proposed in this pull request? This PR proposes to let `WorkerWatcher` run `System.exit` in a separate thread instead of some thread of `RpcEnv`. ### Why are the changes needed? `System.exit` will trigger the shutdown hook to run `executor.stop`, which will result in the same deadlock issue with SPARK-14180. But note that since Spark upgrades to Hadoop 3 recently, each hook now will have a [timeout threshold](https://github.com/apache/hadoop/blob/d4794dd3b2ba365a9d95ad6aafcf43a1ea40f777/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/util/ShutdownHookManager.java#L205-L209) which forcibly interrupt the hook execution once reaches timeout. So, the deadlock issue doesn't really exist in the master branch. However, it's still critical for previous releases and is a wrong behavior that should be fixed. ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? Tested manually. Closes apache#35069 from Ngone51/fix-workerwatcher-exit. Authored-by: yi.wu <[email protected]> Signed-off-by: yi.wu <[email protected]> (cherry picked from commit 639d6f4) Signed-off-by: yi.wu <[email protected]>
catalinii
pushed a commit
to lyft/spark
that referenced
this pull request
Feb 22, 2022
…n a thread out of RpcEnv ### What changes were proposed in this pull request? This PR proposes to let `WorkerWatcher` run `System.exit` in a separate thread instead of some thread of `RpcEnv`. ### Why are the changes needed? `System.exit` will trigger the shutdown hook to run `executor.stop`, which will result in the same deadlock issue with SPARK-14180. But note that since Spark upgrades to Hadoop 3 recently, each hook now will have a [timeout threshold](https://github.com/apache/hadoop/blob/d4794dd3b2ba365a9d95ad6aafcf43a1ea40f777/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/util/ShutdownHookManager.java#L205-L209) which forcibly interrupt the hook execution once reaches timeout. So, the deadlock issue doesn't really exist in the master branch. However, it's still critical for previous releases and is a wrong behavior that should be fixed. ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? Tested manually. Closes apache#35069 from Ngone51/fix-workerwatcher-exit. Authored-by: yi.wu <[email protected]> Signed-off-by: yi.wu <[email protected]> (cherry picked from commit 639d6f4) Signed-off-by: yi.wu <[email protected]>
catalinii
pushed a commit
to lyft/spark
that referenced
this pull request
Mar 4, 2022
…n a thread out of RpcEnv ### What changes were proposed in this pull request? This PR proposes to let `WorkerWatcher` run `System.exit` in a separate thread instead of some thread of `RpcEnv`. ### Why are the changes needed? `System.exit` will trigger the shutdown hook to run `executor.stop`, which will result in the same deadlock issue with SPARK-14180. But note that since Spark upgrades to Hadoop 3 recently, each hook now will have a [timeout threshold](https://github.com/apache/hadoop/blob/d4794dd3b2ba365a9d95ad6aafcf43a1ea40f777/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/util/ShutdownHookManager.java#L205-L209) which forcibly interrupt the hook execution once reaches timeout. So, the deadlock issue doesn't really exist in the master branch. However, it's still critical for previous releases and is a wrong behavior that should be fixed. ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? Tested manually. Closes apache#35069 from Ngone51/fix-workerwatcher-exit. Authored-by: yi.wu <[email protected]> Signed-off-by: yi.wu <[email protected]> (cherry picked from commit 639d6f4) Signed-off-by: yi.wu <[email protected]>
kazuyukitanimura
pushed a commit
to kazuyukitanimura/spark
that referenced
this pull request
Aug 10, 2022
…n a thread out of RpcEnv ### What changes were proposed in this pull request? This PR proposes to let `WorkerWatcher` run `System.exit` in a separate thread instead of some thread of `RpcEnv`. ### Why are the changes needed? `System.exit` will trigger the shutdown hook to run `executor.stop`, which will result in the same deadlock issue with SPARK-14180. But note that since Spark upgrades to Hadoop 3 recently, each hook now will have a [timeout threshold](https://github.com/apache/hadoop/blob/d4794dd3b2ba365a9d95ad6aafcf43a1ea40f777/hadoop-common-project/hadoop-common/src/main/java/org/apache/hadoop/util/ShutdownHookManager.java#L205-L209) which forcibly interrupt the hook execution once reaches timeout. So, the deadlock issue doesn't really exist in the master branch. However, it's still critical for previous releases and is a wrong behavior that should be fixed. ### Does this PR introduce _any_ user-facing change? No. ### How was this patch tested? Tested manually. Closes apache#35069 from Ngone51/fix-workerwatcher-exit. Authored-by: yi.wu <[email protected]> Signed-off-by: yi.wu <[email protected]> (cherry picked from commit 639d6f4) Signed-off-by: yi.wu <[email protected]> (cherry picked from commit 537de84) Signed-off-by: Dongjoon Hyun <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 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?
This PR proposes to let
WorkerWatcherrunSystem.exitin a separate thread instead of some thread ofRpcEnv.Why are the changes needed?
System.exitwill trigger the shutdown hook to runexecutor.stop, which will result in the same deadlock issue with SPARK-14180. But note that since Spark upgrades to Hadoop 3 recently, each hook now will have a timeout threshold which forcibly interrupt the hook execution once reaches timeout. So, the deadlock issue doesn't really exist in the master branch. However, it's still critical for previous releases and is a wrong behavior that should be fixed.Does this PR introduce any user-facing change?
No.
How was this patch tested?
Tested manually.