You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@spark.apache.org by "fred-db (via GitHub)" <gi...@apache.org> on 2024/01/25 13:37:10 UTC
[PR] [SPARK-46861] Avoid Deadlock in DAGScheduler [spark]
fred-db opened a new pull request, #44882:
URL: https://github.com/apache/spark/pull/44882
<!--
Thanks for sending a pull request! Here are some tips for you:
1. If this is your first time, please read our contributor guidelines: https://spark.apache.org/contributing.html
2. Ensure you have added or run the appropriate tests for your PR: https://spark.apache.org/developer-tools.html
3. If the PR is unfinished, add '[WIP]' in your PR title, e.g., '[WIP][SPARK-XXXX] Your PR title ...'.
4. Be sure to keep the PR description updated to reflect all changes.
5. Please write your PR title to summarize what this PR proposes.
6. If possible, provide a concise example to reproduce the issue for a faster review.
7. If you want to add a new configuration, please read the guideline first for naming configurations in
'core/src/main/scala/org/apache/spark/internal/config/ConfigEntry.scala'.
8. If you want to add or modify an error type or message, please read the guideline first in
'core/src/main/resources/error/README.md'.
-->
### What changes were proposed in this pull request?
<!--
Please clarify what changes you are proposing. The purpose of this section is to outline the changes and how this PR fixes the issue.
If possible, please consider writing useful notes for better and faster reviews in your PR. See the examples below.
1. If you refactor some codes with changing classes, showing the class hierarchy will help reviewers.
2. If you fix some SQL features, you can provide some references of other DBMSes.
3. If there is design documentation, please add the link.
4. If there is a discussion in the mailing list, please add the link.
-->
* The DAGScheduler could currently run into a deadlock with another thread if both access the partitions of the same RDD at the same time.
* To make progress in getCacheLocs, we require both exclusive access to the RDD partitions and the location cache. We first lock on the location cache, and then on the RDD.
* When accessing partitions of an RDD, the RDD first acquires exclusive access on the partitions, and then might acquire exclusive access on the location cache.
* If thread 1 is able to acquire access on the RDD, while thread 2 holds the access to the location cache, we can run into a deadlock situation.
* To fix this, acquire locks in the same order. Change the DAGScheduler to first acquire the lock on the RDD, and then the lock on the location cache.
### Why are the changes needed?
<!--
Please clarify why the changes are needed. For instance,
1. If you propose a new API, clarify the use case for a new API.
2. If you fix a bug, you can clarify why it is a bug.
-->
* This is a deadlock you can run into, which can prevent any progress on the cluster.
### Does this PR introduce _any_ user-facing change?
<!--
Note that it means *any* user-facing change including all aspects such as the documentation fix.
If yes, please clarify the previous behavior and the change this PR proposes - provide the console output, description and/or an example to show the behavior difference if possible.
If possible, please also clarify if this is a user-facing change compared to the released Spark versions or within the unreleased branches such as master.
If no, write 'No'.
-->
* No
### How was this patch tested?
<!--
If tests were added, say they were added here. Please make sure to add some test cases that check the changes thoroughly including negative and positive cases if possible.
If it was tested in a way different from regular unit tests, please clarify how you tested step by step, ideally copy and paste-able, so that other reviewers can test and check, and descendants can verify in the future.
If tests were not added, please describe why they were not added and/or why it was difficult to add.
If benchmark tests were added, please run the benchmarks in GitHub Actions for the consistent environment, and the instructions could accord to: https://spark.apache.org/developer-tools.html#github-workflow-benchmarks.
-->
* Unit test that reproduces the issue.
### Was this patch authored or co-authored using generative AI tooling?
<!--
If generative AI tooling has been used in the process of authoring this patch, please include the
phrase: 'Generated-by: ' followed by the name of the tool and its version.
If no, write 'No'.
Please refer to the [ASF Generative Tooling Guidance](https://www.apache.org/legal/generative-tooling.html) for details.
-->
No
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
Re: [PR] [SPARK-46861][CORE] Avoid Deadlock in DAGScheduler [spark]
Posted by "dongjoon-hyun (via GitHub)" <gi...@apache.org>.
dongjoon-hyun commented on PR #44882:
URL: https://github.com/apache/spark/pull/44882#issuecomment-1910578421
Merged to master/3.5/3.4.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
Re: [PR] [SPARK-46861][CORE] Avoid Deadlock in DAGScheduler [spark]
Posted by "mridulm (via GitHub)" <gi...@apache.org>.
mridulm commented on PR #44882:
URL: https://github.com/apache/spark/pull/44882#issuecomment-1912789739
Can you add details of the deadlock stack traces to the jira please ?
I am trying to understand how this is happening on the same RDD.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
Re: [PR] [SPARK-46861][CORE] Avoid Deadlock in DAGScheduler [spark]
Posted by "fred-db (via GitHub)" <gi...@apache.org>.
fred-db commented on PR #44882:
URL: https://github.com/apache/spark/pull/44882#issuecomment-1914299386
Hi @mridulm ,
I'm able to reproduce the deadlock consistently when running the test I added and after removing the call to `stateLock.synchronized` from `getCacheLocs`.
The issue exists essentially with CoalescedRDDs. When calling `getPartitions` on a `CoalescedRDD`, the `PartitionLocations` of the `DefaultPartitionCoalescer` will do a call to `dagScheduler.getPreferredLocs`, which internally calls `getCacheLocs`, causing the deadlock.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
Re: [PR] [SPARK-46861][CORE] Avoid Deadlock in DAGScheduler [spark]
Posted by "dongjoon-hyun (via GitHub)" <gi...@apache.org>.
dongjoon-hyun commented on code in PR #44882:
URL: https://github.com/apache/spark/pull/44882#discussion_r1466635715
##########
core/src/main/scala/org/apache/spark/rdd/RDD.scala:
##########
@@ -224,14 +224,17 @@ abstract class RDD[T: ClassTag](
* not use `this` because RDDs are user-visible, so users might have added their own locking on
* RDDs; sharing that could lead to a deadlock.
*
- * One thread might hold the lock on many of these, for a chain of RDD dependencies; but
- * because DAGs are acyclic, and we only ever hold locks for one path in that DAG, there is no
- * chance of deadlock.
+ * One thread might hold the lock on many of these, for a chain of RDD dependencies. Deadlocks
+ * are possible if we try to lock another resource while holding the stateLock,
+ * and the lock acquisition sequence of these locks is not guaranteed to be the same.
+ * This can lead lead to a deadlock as one thread might first acquire the stateLock,
Review Comment:
nit. `lead lead` -> `lead`
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
Re: [PR] [SPARK-46861][CORE] Avoid Deadlock in DAGScheduler [spark]
Posted by "dongjoon-hyun (via GitHub)" <gi...@apache.org>.
dongjoon-hyun commented on PR #44882:
URL: https://github.com/apache/spark/pull/44882#issuecomment-1913969953
To @mridulm , do you mean the test case is not reproducible from your environment, @mridulm ?
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
Re: [PR] [SPARK-46861][CORE] Avoid Deadlock in DAGScheduler [spark]
Posted by "mridulm (via GitHub)" <gi...@apache.org>.
mridulm commented on PR #44882:
URL: https://github.com/apache/spark/pull/44882#issuecomment-1915407595
@dongjoon-hyun I meant the stack traces indicating a deadlock.
@fred-db thanks for the details - can you update the jira with the traces please ? That will ensure we can reference it in future as well.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
Re: [PR] [SPARK-46861][CORE] Avoid Deadlock in DAGScheduler [spark]
Posted by "mridulm (via GitHub)" <gi...@apache.org>.
mridulm commented on PR #44882:
URL: https://github.com/apache/spark/pull/44882#issuecomment-1913964514
+CC @dongjoon-hyun, since you merged this PR - any insights into how/why this deadlock can occur ? It is not clear to me how this is happening.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
Re: [PR] [SPARK-46861][CORE] Avoid Deadlock in DAGScheduler [spark]
Posted by "dongjoon-hyun (via GitHub)" <gi...@apache.org>.
dongjoon-hyun commented on code in PR #44882:
URL: https://github.com/apache/spark/pull/44882#discussion_r1466638029
##########
core/src/test/scala/org/apache/spark/scheduler/DAGSchedulerSuite.scala:
##########
@@ -612,6 +612,42 @@ class DAGSchedulerSuite extends SparkFunSuite with TempLocalSparkContext with Ti
assertDataStructuresEmpty()
}
+ // Note that this test is NOT perfectly reproducible when there is a deadlock as it uses
+ // Thread.sleep, but it should never fail / flake when there is no deadlock.
+ // If this test starts to flake, this shows that there is a deadlock!
Review Comment:
Got it. Thank you for warning.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org
Re: [PR] [SPARK-46861][CORE] Avoid Deadlock in DAGScheduler [spark]
Posted by "dongjoon-hyun (via GitHub)" <gi...@apache.org>.
dongjoon-hyun closed pull request #44882: [SPARK-46861][CORE] Avoid Deadlock in DAGScheduler
URL: https://github.com/apache/spark/pull/44882
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscribe@spark.apache.org
For additional commands, e-mail: reviews-help@spark.apache.org