[SPARK-50572][SQL] Fix race condition in CachedRDDBuilder.cachedColumnBuffers #49179
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
The previous code had a race condition that mean that `cachedColumnBuffers` could return `null` if another thread was concurrently was calling `clearCache`. The bug is caused by us checking _cachedColumnBuffers and return it as two separate operations outside a synchronized block. So it possible for another thread to set it to `null` after the check but before the return. ``` java.lang.NullPointerException: null at org.apache.spark.sql.execution.columnar.InMemoryTableScanExec.filteredCachedBatches(InMemoryTableScanExec.scala:156) at org.apache.spark.sql.execution.columnar.InMemoryTableScanExec.inputRDD$lzycompute(InMemoryTableScanExec.scala:98) at org.apache.spark.sql.execution.columnar.InMemoryTableScanExec.inputRDD(InMemoryTableScanExec.scala:84) at org.apache.spark.sql.execution.columnar.InMemoryTableScanExec.doExecute(InMemoryTableScanExec.scala:163) at org.apache.spark.sql.execution.SparkPlan.$anonfun$execute$1(SparkPlan.scala:195) at org.apache.spark.sql.execution.SparkPlan.$anonfun$executeQuery$1(SparkPlan.scala:246) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:151) at org.apache.spark.sql.execution.SparkPlan.executeQuery(SparkPlan.scala:243) at org.apache.spark.sql.execution.SparkPlan.execute(SparkPlan.scala:191) at org.apache.spark.sql.execution.InputAdapter.inputRDD(WholeStageCodegenExec.scala:527) at org.apache.spark.sql.execution.InputRDDCodegen.inputRDDs(WholeStageCodegenExec.scala:455) at org.apache.spark.sql.execution.InputRDDCodegen.inputRDDs$(WholeStageCodegenExec.scala:454) at org.apache.spark.sql.execution.InputAdapter.inputRDDs(WholeStageCodegenExec.scala:498) at org.apache.spark.sql.execution.ProjectExec.inputRDDs(basicPhysicalOperators.scala:51) at org.apache.spark.sql.execution.WholeStageCodegenExec.doExecute(WholeStageCodegenExec.scala:751) at org.apache.spark.sql.execution.SparkPlan.$anonfun$execute$1(SparkPlan.scala:195) at org.apache.spark.sql.execution.SparkPlan.$anonfun$executeQuery$1(SparkPlan.scala:246) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:151) at org.apache.spark.sql.execution.SparkPlan.executeQuery(SparkPlan.scala:243) at org.apache.spark.sql.execution.SparkPlan.execute(SparkPlan.scala:191) at org.apache.spark.sql.execution.SparkPlan.getByteArrayRdd(SparkPlan.scala:364) at org.apache.spark.sql.execution.SparkPlan.executeCollect(SparkPlan.scala:445) at org.apache.spark.sql.Dataset.collectFromPlan(Dataset.scala:4218) at org.apache.spark.sql.Dataset.$anonfun$collect$1(Dataset.scala:3459) at org.apache.spark.sql.Dataset.$anonfun$withAction$2(Dataset.scala:4208) at org.apache.spark.sql.execution.QueryExecution$.withInternalError(QueryExecution.scala:526) ... 23 more ```
Contributor
Author
|
@cloud-fan since you were involved in #21018 where the problematic code was introduced. |
Contributor
Author
|
An alternative fix could be to mark the entire method as synchronized if we do not think these are on a critical path. |
|
We're closing this PR because it hasn't been updated in a while. This isn't a judgement on the merit of the PR in any way. It's just a way of keeping the PR queue manageable. |
|
I'd like to have a look at it, when I have time. |
Contributor
Author
It was just waiting for someone to review it. As far as I am aware it good to go as it is and it fixes a real race condition. |
|
I find this PR quite interesting indeed. It's a pity this hasn't gone ahead. Did you send a message to the dev mail list? |
Contributor
Author
|
Sorry about the late response, but no I don't think I did. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 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?
Fix race condition in the class
CachedRDDBuilder.Why are the changes needed?
The previous code had a race condition that mean that
cachedColumnBufferscould returnnullif another thread was concurrently was callingclearCache.The bug is caused by us checking
_cachedColumnBuffersand return it as two separate operations outside a synchronized block. So it possible for another thread to set it tonullafter the check but before the return.Does this PR introduce any user-facing change?
Yes, fixes a race condition that can cause crashes.
How was this patch tested?
Adds a test that shows that reliably fails on the old code. Not sure if want to merge that style of test as it very specific for this bug and takes 3 seconds to run.
Was this patch authored or co-authored using generative AI tooling?
No.