You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@spark.apache.org by "cloud-fan (via GitHub)" <gi...@apache.org> on 2024/03/01 10:36:20 UTC
[PR] [SPARK-47247][SQL] Use smaller target size when coalescing partitions with exploding joins [spark]
cloud-fan opened a new pull request, #45357:
URL: https://github.com/apache/spark/pull/45357
<!--
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
'common/utils/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.
-->
This PR changes the target partition size of AQE partition coalescing from `spark.sql.adaptive.advisoryPartitionSizeInBytes` (default 64MB) to `spark.sql.adaptive.coalescePartitions.minPartitionSize` (default 1MB) for non-equi joins, namely, broadcast nested loop join and cartesian product join, in order to minimize OOM risks as these join operators tend to be exploding joins and usually work better with smaller partitions compared to other operators.
### 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.
-->
reduce the OOM risk, as after data exploding, 64mb input can become super large
### 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.
-->
new test
### 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-47247][SQL] Use smaller target size when coalescing partitions with exploding joins [spark]
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #45357:
URL: https://github.com/apache/spark/pull/45357#discussion_r1510594245
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/adaptive/CoalesceShufflePartitions.scala:
##########
@@ -126,9 +126,12 @@ case class CoalesceShufflePartitions(session: SparkSession) extends AQEShuffleRe
// data sources may request a particular advisory partition size for the final write stage
// if it happens, the advisory partition size will be set in ShuffleQueryStageExec
// only one shuffle stage is expected in such cases
- private def advisoryPartitionSize(shuffleStages: Seq[ShuffleStageInfo]): Long = {
+ private def advisoryPartitionSize(coalesceGroup: CoalesceGroup): Long = {
+ if (coalesceGroup.hasExplodingJoin) {
+ return conf.getConf(SQLConf.COALESCE_PARTITIONS_MIN_PARTITION_SIZE)
Review Comment:
Then we probably need to add reduce ratio which needs more time to test and build it correctly: https://github.com/apache/spark/pull/45357#pullrequestreview-1911078774
By having the reduce ratio, we can do coalescing differently for different coalesce groups. Maybe we can do it as a follow-up.
--
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-47247][SQL] Use smaller target size when coalescing partitions with exploding joins [spark]
Posted by "yaooqinn (via GitHub)" <gi...@apache.org>.
yaooqinn commented on code in PR #45357:
URL: https://github.com/apache/spark/pull/45357#discussion_r1512030230
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/adaptive/CoalesceShufflePartitions.scala:
##########
@@ -126,9 +126,12 @@ case class CoalesceShufflePartitions(session: SparkSession) extends AQEShuffleRe
// data sources may request a particular advisory partition size for the final write stage
// if it happens, the advisory partition size will be set in ShuffleQueryStageExec
// only one shuffle stage is expected in such cases
- private def advisoryPartitionSize(shuffleStages: Seq[ShuffleStageInfo]): Long = {
+ private def advisoryPartitionSize(coalesceGroup: CoalesceGroup): Long = {
+ if (coalesceGroup.hasExplodingJoin) {
+ return conf.getConf(SQLConf.COALESCE_PARTITIONS_MIN_PARTITION_SIZE)
Review Comment:
Maybe we can respect `spark.sql.adaptive.coalescePartitions.parallelismFirst here`?
--
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-47247][SQL] Use smaller target size when coalescing partitions with exploding joins [spark]
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on code in PR #45357:
URL: https://github.com/apache/spark/pull/45357#discussion_r1510518917
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/adaptive/CoalesceShufflePartitions.scala:
##########
@@ -126,9 +126,12 @@ case class CoalesceShufflePartitions(session: SparkSession) extends AQEShuffleRe
// data sources may request a particular advisory partition size for the final write stage
// if it happens, the advisory partition size will be set in ShuffleQueryStageExec
// only one shuffle stage is expected in such cases
- private def advisoryPartitionSize(shuffleStages: Seq[ShuffleStageInfo]): Long = {
+ private def advisoryPartitionSize(coalesceGroup: CoalesceGroup): Long = {
+ if (coalesceGroup.hasExplodingJoin) {
+ return conf.getConf(SQLConf.COALESCE_PARTITIONS_MIN_PARTITION_SIZE)
Review Comment:
shall we add a new config like `explodingStageAdvisorySizeFactor` to make the advisory size easy to change ?
--
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-47247][SQL] Use smaller target size when coalescing partitions with exploding joins [spark]
Posted by "ulysses-you (via GitHub)" <gi...@apache.org>.
ulysses-you commented on code in PR #45357:
URL: https://github.com/apache/spark/pull/45357#discussion_r1510571235
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/adaptive/CoalesceShufflePartitions.scala:
##########
@@ -126,9 +126,12 @@ case class CoalesceShufflePartitions(session: SparkSession) extends AQEShuffleRe
// data sources may request a particular advisory partition size for the final write stage
// if it happens, the advisory partition size will be set in ShuffleQueryStageExec
// only one shuffle stage is expected in such cases
- private def advisoryPartitionSize(shuffleStages: Seq[ShuffleStageInfo]): Long = {
+ private def advisoryPartitionSize(coalesceGroup: CoalesceGroup): Long = {
+ if (coalesceGroup.hasExplodingJoin) {
+ return conf.getConf(SQLConf.COALESCE_PARTITIONS_MIN_PARTITION_SIZE)
Review Comment:
I thought if the query contain multi-stages and there are more than one stage contain exploding operators. People may want to change this config more fine-grained. That's why I tried to support plan fragment config before.
I'm fine to use the `COALESCE_PARTITIONS_MIN_PARTITION_SIZE`.
--
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-47247][SQL] Use smaller target size when coalescing partitions with exploding joins [spark]
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on PR #45357:
URL: https://github.com/apache/spark/pull/45357#issuecomment-1973167129
fine-grained reduce ratio looks nice, but it needs a lot of tuning, and more or less based on experience. I don't think fixed reduce ratios work for all the cases and we probably need to add many configs. The approach in this PR is like a coarse-grained fallback: if there are exploding joins, do not coalesce (since the default is 1mb) but people can still tune it a bit.
`ExpandExec` should not have many projectLists and won't explode the data too much. But I don't know how people use it in real world use cases.
--
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-47247][SQL] Use smaller target size when coalescing partitions with exploding joins [spark]
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on PR #45357:
URL: https://github.com/apache/spark/pull/45357#issuecomment-1972942423
cc @yaooqinn @ulysses-you
--
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-47247][SQL] Use smaller target size when coalescing partitions with exploding joins [spark]
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #45357:
URL: https://github.com/apache/spark/pull/45357#discussion_r1510551047
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/adaptive/CoalesceShufflePartitions.scala:
##########
@@ -126,9 +126,12 @@ case class CoalesceShufflePartitions(session: SparkSession) extends AQEShuffleRe
// data sources may request a particular advisory partition size for the final write stage
// if it happens, the advisory partition size will be set in ShuffleQueryStageExec
// only one shuffle stage is expected in such cases
- private def advisoryPartitionSize(shuffleStages: Seq[ShuffleStageInfo]): Long = {
+ private def advisoryPartitionSize(coalesceGroup: CoalesceGroup): Long = {
+ if (coalesceGroup.hasExplodingJoin) {
+ return conf.getConf(SQLConf.COALESCE_PARTITIONS_MIN_PARTITION_SIZE)
Review Comment:
isn't it harder to tune two configs (advisory size + advisory size factor for exploding) than just one advisory size for exploding? I didn't add a new config as I think `COALESCE_PARTITIONS_MIN_PARTITION_SIZE` has a reasonable default value to turn off coalesce for this case, and we do need to respect it when coalescing partitions.
--
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-47247][SQL] Use smaller target size when coalescing partitions with exploding joins [spark]
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan closed pull request #45357: [SPARK-47247][SQL] Use smaller target size when coalescing partitions with exploding joins
URL: https://github.com/apache/spark/pull/45357
--
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-47247][SQL] Use smaller target size when coalescing partitions with exploding joins [spark]
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on PR #45357:
URL: https://github.com/apache/spark/pull/45357#issuecomment-1979976679
thanks for the review, merging to master!
--
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-47247][SQL] Use smaller target size when coalescing partitions with exploding joins [spark]
Posted by "cloud-fan (via GitHub)" <gi...@apache.org>.
cloud-fan commented on code in PR #45357:
URL: https://github.com/apache/spark/pull/45357#discussion_r1512860006
##########
sql/core/src/main/scala/org/apache/spark/sql/execution/adaptive/CoalesceShufflePartitions.scala:
##########
@@ -126,9 +126,12 @@ case class CoalesceShufflePartitions(session: SparkSession) extends AQEShuffleRe
// data sources may request a particular advisory partition size for the final write stage
// if it happens, the advisory partition size will be set in ShuffleQueryStageExec
// only one shuffle stage is expected in such cases
- private def advisoryPartitionSize(shuffleStages: Seq[ShuffleStageInfo]): Long = {
+ private def advisoryPartitionSize(coalesceGroup: CoalesceGroup): Long = {
+ if (coalesceGroup.hasExplodingJoin) {
+ return conf.getConf(SQLConf.COALESCE_PARTITIONS_MIN_PARTITION_SIZE)
Review Comment:
`parallelismFirst` affects the `minNumPartitions`, seems to be a different thing here.
--
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