[SPARK-33119][SQL] ScalarSubquery should returns the first two rows to avoid Driver OOM #30016
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
|
Kubernetes integration test starting |
|
Kubernetes integration test status success |
maropu
approved these changes
Oct 12, 2020
|
Test build #129683 has finished for PR 30016 at commit
|
HyukjinKwon
approved these changes
Oct 13, 2020
Member
|
cc @cloud-fan and @maryannxue FYI. Merged to master. |
HyukjinKwon
pushed a commit
that referenced
this pull request
Dec 15, 2020
### What changes were proposed in this pull request? If we call `SubqueryExec.executeTake`, it will call `SubqueryExec.execute` which will trigger the codegen of the query plan and create an RDD. However, `SubqueryExec` already has a thread (`SubqueryExec.relationFuture`) to execute the query plan, which means we have 2 threads triggering codegen of the same query plan at the same time. Spark codegen is not thread-safe, as we have places like `HashAggregateExec.bufferVars` that is a shared variable. The bug in `SubqueryExec` may lead to correctness bugs. Since https://issues.apache.org/jira/browse/SPARK-33119, `ScalarSubquery` will call `SubqueryExec.executeTake`, so flaky tests start to appear. This PR fixes the bug by reimplementing #30016 . We should pass the number of rows we want to collect to `SubqueryExec` at planning time, so that we can use `executeTake` inside `SubqueryExec.relationFuture`, and the caller side should always call `SubqueryExec.executeCollect`. This PR also adds checks so that we can make sure only `SubqueryExec.executeCollect` is called. ### Why are the changes needed? fix correctness bug. ### Does this PR introduce _any_ user-facing change? no ### How was this patch tested? run `build/sbt "sql/testOnly *SQLQueryTestSuite -- -z scalar-subquery-select"` more than 10 times. Previously it fails, now it passes. Closes #30765 from cloud-fan/bug. Authored-by: Wenchen Fan <[email protected]> Signed-off-by: HyukjinKwon <[email protected]>
HyukjinKwon
pushed a commit
that referenced
this pull request
Dec 15, 2020
### What changes were proposed in this pull request? If we call `SubqueryExec.executeTake`, it will call `SubqueryExec.execute` which will trigger the codegen of the query plan and create an RDD. However, `SubqueryExec` already has a thread (`SubqueryExec.relationFuture`) to execute the query plan, which means we have 2 threads triggering codegen of the same query plan at the same time. Spark codegen is not thread-safe, as we have places like `HashAggregateExec.bufferVars` that is a shared variable. The bug in `SubqueryExec` may lead to correctness bugs. Since https://issues.apache.org/jira/browse/SPARK-33119, `ScalarSubquery` will call `SubqueryExec.executeTake`, so flaky tests start to appear. This PR fixes the bug by reimplementing #30016 . We should pass the number of rows we want to collect to `SubqueryExec` at planning time, so that we can use `executeTake` inside `SubqueryExec.relationFuture`, and the caller side should always call `SubqueryExec.executeCollect`. This PR also adds checks so that we can make sure only `SubqueryExec.executeCollect` is called. ### Why are the changes needed? fix correctness bug. ### Does this PR introduce _any_ user-facing change? no ### How was this patch tested? run `build/sbt "sql/testOnly *SQLQueryTestSuite -- -z scalar-subquery-select"` more than 10 times. Previously it fails, now it passes. Closes #30765 from cloud-fan/bug. Authored-by: Wenchen Fan <[email protected]> Signed-off-by: HyukjinKwon <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 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?
ScalarSubqueryshould returns the first two rows.Why are the changes needed?
To avoid Driver OOM.
Does this PR introduce any user-facing change?
No.
How was this patch tested?
Existing test:
spark/sql/core/src/test/scala/org/apache/spark/sql/SubquerySuite.scala
Lines 147 to 154 in d6f3138