Skip to content

Commit

Permalink
[SPARK-39551][SQL] Add AQE invalid plan check
Browse files Browse the repository at this point in the history
This PR adds a check for invalid plans in AQE replanning process. The check will throw exceptions when it detects an invalid plan, causing AQE to void the current replanning result and keep using the latest valid plan.

AQE logical optimization rules can lead to invalid physical plans and cause runtime exceptions as certain physical plan nodes are not compatible with others. E.g., `BroadcastExchangeExec` can only work as a direct child of broadcast join nodes, but it could appear under other incompatible physical plan nodes because of empty relation propagation.

No.

Added UT.

Closes apache#36953 from maryannxue/validate-aqe.

Authored-by: Maryann Xue <maryann.xue@gmail.com>
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
(cherry picked from commit 58b91b1)
Signed-off-by: Wenchen Fan <wenchen@databricks.com>
(cherry picked from commit 3cf3048)
  • Loading branch information
maryannxue authored and lwz9103 committed Jun 5, 2024
1 parent c90acc1 commit 17b1db1
Show file tree
Hide file tree
Showing 4 changed files with 163 additions and 32 deletions.
Original file line number Diff line number Diff line change
Expand Up @@ -118,6 +118,7 @@ case class AdaptiveSparkPlanExec(
Seq(
RemoveRedundantProjects,
ensureRequirements,
ValidateSparkPlan,
ReplaceHashWithSortAgg,
RemoveRedundantSorts,
DisableUnnecessaryBucketedScan,
Expand Down Expand Up @@ -322,16 +323,19 @@ case class AdaptiveSparkPlanExec(
// plans are updated, we can clear the query stage list because at this point the two plans
// are semantically and physically in sync again.
val logicalPlan = replaceWithQueryStagesInLogicalPlan(currentLogicalPlan, stagesToReplace)
val (newPhysicalPlan, newLogicalPlan) = reOptimize(logicalPlan)
val origCost = costEvaluator.evaluateCost(currentPhysicalPlan)
val newCost = costEvaluator.evaluateCost(newPhysicalPlan)
if (newCost < origCost ||
val afterReOptimize = reOptimize(logicalPlan)
if (afterReOptimize.isDefined) {
val (newPhysicalPlan, newLogicalPlan) = afterReOptimize.get
val origCost = costEvaluator.evaluateCost(currentPhysicalPlan)
val newCost = costEvaluator.evaluateCost(newPhysicalPlan)
if (newCost < origCost ||
(newCost == origCost && currentPhysicalPlan != newPhysicalPlan)) {
logOnLevel(s"Plan changed from $currentPhysicalPlan to $newPhysicalPlan")
cleanUpTempTags(newPhysicalPlan)
currentPhysicalPlan = newPhysicalPlan
currentLogicalPlan = newLogicalPlan
stagesToReplace = Seq.empty[QueryStageExec]
logOnLevel(s"Plan changed from $currentPhysicalPlan to $newPhysicalPlan")
cleanUpTempTags(newPhysicalPlan)
currentPhysicalPlan = newPhysicalPlan
currentLogicalPlan = newLogicalPlan
stagesToReplace = Seq.empty[QueryStageExec]
}
}
// Now that some stages have finished, we can try creating new stages.
result = createQueryStages(currentPhysicalPlan)
Expand Down Expand Up @@ -661,29 +665,35 @@ case class AdaptiveSparkPlanExec(
/**
* Re-optimize and run physical planning on the current logical plan based on the latest stats.
*/
private def reOptimize(logicalPlan: LogicalPlan): (SparkPlan, LogicalPlan) = {
logicalPlan.invalidateStatsCache()
val optimized = optimizer.execute(logicalPlan)
val sparkPlan = context.session.sessionState.planner.plan(ReturnAnswer(optimized)).next()
val newPlan = applyPhysicalRules(
sparkPlan,
preprocessingRules ++ queryStagePreparationRules,
Some((planChangeLogger, "AQE Replanning")))

// When both enabling AQE and DPP, `PlanAdaptiveDynamicPruningFilters` rule will
// add the `BroadcastExchangeExec` node manually in the DPP subquery,
// not through `EnsureRequirements` rule. Therefore, when the DPP subquery is complicated
// and need to be re-optimized, AQE also need to manually insert the `BroadcastExchangeExec`
// node to prevent the loss of the `BroadcastExchangeExec` node in DPP subquery.
// Here, we also need to avoid to insert the `BroadcastExchangeExec` node when the newPlan
// is already the `BroadcastExchangeExec` plan after apply the `LogicalQueryStageStrategy` rule.
val finalPlan = currentPhysicalPlan match {
case b: BroadcastExchangeLike
if (!newPlan.isInstanceOf[BroadcastExchangeLike]) => b.withNewChildren(Seq(newPlan))
case _ => newPlan
}
private def reOptimize(logicalPlan: LogicalPlan): Option[(SparkPlan, LogicalPlan)] = {
try {
logicalPlan.invalidateStatsCache()
val optimized = optimizer.execute(logicalPlan)
val sparkPlan = context.session.sessionState.planner.plan(ReturnAnswer(optimized)).next()
val newPlan = applyPhysicalRules(
sparkPlan,
preprocessingRules ++ queryStagePreparationRules,
Some((planChangeLogger, "AQE Replanning")))

// When both enabling AQE and DPP, `PlanAdaptiveDynamicPruningFilters` rule will
// add the `BroadcastExchangeExec` node manually in the DPP subquery,
// not through `EnsureRequirements` rule. Therefore, when the DPP subquery is complicated
// and need to be re-optimized, AQE also need to manually insert the `BroadcastExchangeExec`
// node to prevent the loss of the `BroadcastExchangeExec` node in DPP subquery.
// Here, we also need to avoid to insert the `BroadcastExchangeExec` node when the newPlan is
// already the `BroadcastExchangeExec` plan after apply the `LogicalQueryStageStrategy` rule.
val finalPlan = currentPhysicalPlan match {
case b: BroadcastExchangeLike
if (!newPlan.isInstanceOf[BroadcastExchangeLike]) => b.withNewChildren(Seq(newPlan))
case _ => newPlan
}

(finalPlan, optimized)
Some((finalPlan, optimized))
} catch {
case e: InvalidAQEPlanException[_] =>
logOnLevel(s"Re-optimize - ${e.getMessage()}:\n${e.plan}")
None
}
}

/**
Expand Down
Original file line number Diff line number Diff line change
@@ -0,0 +1,30 @@
/*
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/

package org.apache.spark.sql.execution.adaptive

import org.apache.spark.sql.catalyst.plans.QueryPlan

/**
* Exception thrown when an invalid query plan is detected in AQE replanning,
* in which case AQE will stop the current replanning process and keep using the latest valid plan.
*
* @param message The reason why the plan is considered invalid.
* @param plan The invalid plan/sub-plan.
*/
case class InvalidAQEPlanException[QueryType <: QueryPlan[_]](message: String, plan: QueryType)
extends Exception(message)
Original file line number Diff line number Diff line change
@@ -0,0 +1,68 @@
/*
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/

package org.apache.spark.sql.execution.adaptive

import org.apache.spark.sql.catalyst.optimizer.{BuildLeft, BuildRight}
import org.apache.spark.sql.catalyst.rules.Rule
import org.apache.spark.sql.execution.SparkPlan
import org.apache.spark.sql.execution.joins.{BroadcastHashJoinExec, BroadcastNestedLoopJoinExec}

/**
* Detects invalid physical plans generated by AQE replanning and throws `InvalidAQEPlanException`
* if such plans are detected. This rule should be called after EnsureRequirements where all
* necessary Exchange nodes are added.
*/
object ValidateSparkPlan extends Rule[SparkPlan] {

def apply(plan: SparkPlan): SparkPlan = {
validate(plan)
plan
}

/**
* Validate that the plan satisfies the following condition:
* - BroadcastQueryStage only appears as the immediate child and the build side of a broadcast
* hash join or broadcast nested loop join.
*/
private def validate(plan: SparkPlan): Unit = plan match {
case b: BroadcastHashJoinExec =>
val (buildPlan, probePlan) = b.buildSide match {
case BuildLeft => (b.left, b.right)
case BuildRight => (b.right, b.left)
}
if (!buildPlan.isInstanceOf[BroadcastQueryStageExec]) {
validate(buildPlan)
}
validate(probePlan)
case b: BroadcastNestedLoopJoinExec =>
val (buildPlan, probePlan) = b.buildSide match {
case BuildLeft => (b.left, b.right)
case BuildRight => (b.right, b.left)
}
if (!buildPlan.isInstanceOf[BroadcastQueryStageExec]) {
validate(buildPlan)
}
validate(probePlan)
case q: BroadcastQueryStageExec => errorOnInvalidBroadcastQueryStage(q)
case _ => plan.children.foreach(validate)
}

private def errorOnInvalidBroadcastQueryStage(plan: SparkPlan): Unit = {
throw InvalidAQEPlanException("Invalid broadcast query stage", plan)
}
}
Original file line number Diff line number Diff line change
Expand Up @@ -33,7 +33,7 @@ import org.apache.spark.sql.execution.command.DataWritingCommandExec
import org.apache.spark.sql.execution.datasources.noop.NoopDataSource
import org.apache.spark.sql.execution.datasources.v2.V2TableWriteExec
import org.apache.spark.sql.execution.exchange.{BroadcastExchangeExec, ENSURE_REQUIREMENTS, Exchange, REPARTITION_BY_COL, REPARTITION_BY_NUM, ReusedExchangeExec, ShuffleExchangeExec, ShuffleExchangeLike, ShuffleOrigin}
import org.apache.spark.sql.execution.joins.{BaseJoinExec, BroadcastHashJoinExec, ShuffledHashJoinExec, ShuffledJoin, SortMergeJoinExec}
import org.apache.spark.sql.execution.joins.{BaseJoinExec, BroadcastHashJoinExec, BroadcastNestedLoopJoinExec, ShuffledHashJoinExec, ShuffledJoin, SortMergeJoinExec}
import org.apache.spark.sql.execution.metric.SQLShuffleReadMetricsReporter
import org.apache.spark.sql.execution.ui.SparkListenerSQLAdaptiveExecutionUpdate
import org.apache.spark.sql.functions._
Expand Down Expand Up @@ -103,6 +103,12 @@ class AdaptiveQueryExecSuite
}
}

def findTopLevelBroadcastNestedLoopJoin(plan: SparkPlan): Seq[BaseJoinExec] = {
collect(plan) {
case j: BroadcastNestedLoopJoinExec => j
}
}

private def findTopLevelSortMergeJoin(plan: SparkPlan): Seq[SortMergeJoinExec] = {
collect(plan) {
case j: SortMergeJoinExec => j
Expand Down Expand Up @@ -2576,6 +2582,23 @@ class AdaptiveQueryExecSuite
assert(findTopLevelAggregate(adaptive5).size == 4)
}
}

test("SPARK-39551: Invalid plan check - invalid broadcast query stage") {
withSQLConf(
SQLConf.ADAPTIVE_EXECUTION_ENABLED.key -> "true") {
val (_, adaptivePlan) = runAdaptiveAndVerifyResult(
"""
|SELECT /*+ BROADCAST(t3) */ t3.b, count(t3.a) FROM testData2 t1
|INNER JOIN testData2 t2
|ON t1.b = t2.b AND t1.a = 0
|RIGHT OUTER JOIN testData2 t3
|ON t1.a > t3.a
|GROUP BY t3.b
""".stripMargin
)
assert(findTopLevelBroadcastNestedLoopJoin(adaptivePlan).size == 1)
}
}
}

/**
Expand Down

0 comments on commit 17b1db1

Please sign in to comment.