taiyang-li opened a new pull request, #11491:
URL: https://github.com/apache/incubator-gluten/pull/11491
## Summary
Extend the stage-level resource profile adjustment rule and its coverage so
that stages containing PartialProject/PartialGenerate operators can also
benefit from heap/off-heap tuning.
## Changes
- **GlutenAutoAdjustStageResourceProfile**
- Compute `fallenNodeRatio` once and introduce a new `hasPartialNode` flag
based on the stage plan nodes.
- Treat stages that contain PartialProject/PartialGenerate operators as
candidates for the "partial fallback" resource adjustment path, even when the
fallen-node ratio is below the configured threshold and the stage is not a
whole-stage fallback.
- Keep using existing experimental knobs `autoAdjustStageRPHeapRatio` and
`autoAdjustStageRPOffHeapRatio` to tune heap and off-heap memory, and maintain
the existing behavior for whole-stage fallback.
- Add a helper `hasPartialOperator(planNodes: Seq[SparkPlan])` in the
companion object, implemented via `SparkPlan.nodeName` string matching to avoid
compile-time coupling with backend-specific exec classes.
- Add log output when the partial-fallback/partial-operator path is taken
to aid observability.
- **Tests (Velox backend)**
- Extend `AutoAdjustStageResourceProfileSuite` with helpers to count
`ColumnarPartialProjectExec` / `ColumnarPartialGenerateExec` and manage
temporary Hive functions.
- Add a test that enables `ENABLE_COLUMNAR_PARTIAL_PROJECT` and runs a
grouped aggregation with a Scala UDF on `tmp1`, asserting that the executed
plan contains `ColumnarPartialProjectExec` and exactly one
`ApplyResourceProfileExec`.
- Add a test that enables `ENABLE_COLUMNAR_PARTIAL_GENERATE`, registers a
`SimpleUDTF`-based Hive UDTF, and runs a query with `lateral view` and a
subsequent aggregation on `tmp1`, asserting that the executed plan contains
`ColumnarPartialGenerateExec` and exactly one `ApplyResourceProfileExec`.
- **Documentation**
- Update `docs/get-started/VeloxStageResourceAdj.md`:
- Add `spark.gluten.auto.adjustStageResources.offheap.ratio` to the
configuration table with description and default value, and fix the example key
for enabling auto adjustment to `spark.gluten.auto.adjustStageResource.enabled`.
- Introduce a new **Scenario 2: PartialProject / PartialGenerate
Exists** subsection describing when and how stages with
PartialProject/PartialGenerate trigger moderate heap/off-heap adjustments using
the existing ratios.
- Extend the verification section with guidance on checking executor JVM
options (e.g. via `ps aux | grep java`) to validate that stage-specific
heap/off-heap settings take effect.
## Testing
- `mvn -q -DskipTests install` (from project root) to build `gluten-core`,
`gluten-substrait`, `gluten-ras`, `shims`, and related modules: **PASS** after
fixing spotless violations in `GlutenAutoAdjustStageResourceProfile.scala`.
- `cd backends-velox && mvn -q -Dtest=AutoAdjustStageResourceProfileSuite
test`: **FAILED** due to unresolved dependency
`org.apache.gluten:gluten-arrow:jar:1.6.0-SNAPSHOT` (Arrow 15.0.0-gluten
artifacts are not available in the configured Maven mirror in this
environment). The new tests compile, but the suite could not be executed
end-to-end without the Arrow artifacts.
Change-Id: I91541b89c128c6f8e8751440ed1fd5606a3d7049
<!--
Thank you for submitting a pull request! Here are some tips:
1. For first-time contributors, please read our contributing guide:
https://github.com/apache/incubator-gluten/blob/main/CONTRIBUTING.md
2. If necessary, create a GitHub issue for discussion beforehand to avoid
duplicate work.
3. If the PR is specific to a single backend, include [VL] or [CH] in the PR
title to indicate the
Velox or ClickHouse backend, respectively.
4. If the PR is not ready for review, please mark it as a draft.
-->
## What changes are proposed in this pull request?
<!--
Provide a clear and concise description of the changes introduced in this PR.
Ensure the PR description aligns with the code changes, especially after
updates.
If applicable, include "Fixes #<GitHub_Issue_ID>" to automatically close the
corresponding issue
when the PR is merged.
-->
## How was this patch tested?
<!--
Describe how the changes were tested, if applicable.
Include new tests to validate the functionality, if necessary.
For UI-related changes, attach screenshots to demonstrate the updates.
-->
--
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: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
