You are viewing a plain text version of this content. The canonical link for it is here.
Posted to reviews@spark.apache.org by GitBox <gi...@apache.org> on 2022/03/04 06:16:06 UTC
[GitHub] [spark] HeartSaVioR commented on a change in pull request #35673: [SPARK-38204][SS] Use StatefulOpClusteredDistribution for stateful operators with respecting backward compatibility
HeartSaVioR commented on a change in pull request #35673:
URL: https://github.com/apache/spark/pull/35673#discussion_r819293042
##########
File path: sql/core/src/test/scala/org/apache/spark/sql/streaming/StreamingSessionWindowDistributionSuite.scala
##########
@@ -0,0 +1,232 @@
+/*
+ * 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.streaming
+
+import java.io.File
+
+import org.apache.commons.io.FileUtils
+
+import org.apache.spark.internal.Logging
+import org.apache.spark.sql.catalyst.plans.physical.UnspecifiedDistribution
+import org.apache.spark.sql.execution.aggregate.BaseAggregateExec
+import org.apache.spark.sql.execution.streaming.{MemoryStream, SessionWindowStateStoreRestoreExec, SessionWindowStateStoreSaveExec}
+import org.apache.spark.sql.functions.{count, session_window}
+import org.apache.spark.sql.internal.SQLConf
+import org.apache.spark.sql.streaming.util.StatefulOpClusteredDistributionTestHelper
+import org.apache.spark.util.Utils
+
+class StreamingSessionWindowDistributionSuite extends StreamTest
+ with StatefulOpClusteredDistributionTestHelper with Logging {
+
+ import testImplicits._
+
+ test("SPARK-38204: session window aggregation should require StatefulOpClusteredDistribution " +
+ "from children") {
+
+ withSQLConf(
+ // exclude partial merging session to simplify test
+ SQLConf.STREAMING_SESSION_WINDOW_MERGE_SESSIONS_IN_LOCAL_PARTITION.key -> "false") {
+
+ val inputData = MemoryStream[(String, String, Long)]
+
+ // Split the lines into words, treat words as sessionId of events
+ val events = inputData.toDF()
+ .select($"_1".as("value"), $"_2".as("userId"), $"_3".as("timestamp"))
+ .withColumn("eventTime", $"timestamp".cast("timestamp"))
+ .withWatermark("eventTime", "30 seconds")
+ .selectExpr("explode(split(value, ' ')) AS sessionId", "userId", "eventTime")
+
+ val sessionUpdates = events
+ .repartition($"userId")
+ .groupBy(session_window($"eventTime", "10 seconds") as 'session, 'sessionId, 'userId)
+ .agg(count("*").as("numEvents"))
+ .selectExpr("sessionId", "userId", "CAST(session.start AS LONG)",
+ "CAST(session.end AS LONG)",
+ "CAST(session.end AS LONG) - CAST(session.start AS LONG) AS durationMs",
+ "numEvents")
+
+ testStream(sessionUpdates, OutputMode.Append())(
+ AddData(inputData,
+ ("hello world spark streaming", "key1", 40L),
+ ("world hello structured streaming", "key2", 41L)
+ ),
+
+ // skip checking the result, since we focus to verify the physical plan
+ ProcessAllAvailable(),
+ Execute { query =>
+ val numPartitions = query.lastExecution.numStateStores
+
+ val operators = query.lastExecution.executedPlan.collect {
+ case s: SessionWindowStateStoreRestoreExec => s
+ case s: SessionWindowStateStoreSaveExec => s
+ }
+
+ assert(operators.nonEmpty)
+ operators.foreach { stateOp =>
+ assert(requireStatefulOpClusteredDistribution(stateOp, Seq(Seq("sessionId", "userId")),
+ numPartitions))
+ assert(hasDesiredHashPartitioningInChildren(stateOp, Seq(Seq("sessionId", "userId")),
+ numPartitions))
+ }
+
+ // Verify aggregations in between, except partial aggregation.
+ // This includes MergingSessionsExec.
+ val allAggregateExecs = query.lastExecution.executedPlan.collect {
+ case a: BaseAggregateExec => a
+ }
+
+ val aggregateExecsWithoutPartialAgg = allAggregateExecs.filter {
+ _.requiredChildDistribution.head != UnspecifiedDistribution
+ }
+
+ // We expect single partial aggregation since we disable partial merging sessions.
+ // Remaining agg execs should have child producing expected output partitioning.
+ assert(allAggregateExecs.length - 1 === aggregateExecsWithoutPartialAgg.length)
+
+ // For aggregate execs, we make sure output partitioning of the children is same as
+ // we expect, HashPartitioning with clustering keys & number of partitions.
+ aggregateExecsWithoutPartialAgg.foreach { aggr =>
+ assert(hasDesiredHashPartitioningInChildren(aggr, Seq(Seq("sessionId", "userId")),
+ numPartitions))
+ }
+ }
+ )
+ }
+ }
+
+ test("SPARK-38204: session window aggregation should require ClusteredDistribution " +
+ "from children if the query starts from checkpoint in 3.2") {
+
+ withSQLConf(
+ // exclude partial merging session to simplify test
+ SQLConf.STREAMING_SESSION_WINDOW_MERGE_SESSIONS_IN_LOCAL_PARTITION.key -> "false",
+ // explode(split(value, ' ')) with value as nullable field seems to give different
+ // nullability between Spark 3.2 and 3.3 (nullable in 3.2 and non-null in 3.3)
+ // This should be allowed, but unfortunately, nullability check in state store schema
+ // checker seems to have a bug (`from` and `to` are swapped). Applying temporary
+ // workaround for now.
+ // TODO: investigate the schema checker and remove this once it is resolved.
Review comment:
https://github.com/apache/spark/pull/35731
##########
File path: sql/core/src/test/scala/org/apache/spark/sql/streaming/StreamingSessionWindowDistributionSuite.scala
##########
@@ -0,0 +1,232 @@
+/*
+ * 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.streaming
+
+import java.io.File
+
+import org.apache.commons.io.FileUtils
+
+import org.apache.spark.internal.Logging
+import org.apache.spark.sql.catalyst.plans.physical.UnspecifiedDistribution
+import org.apache.spark.sql.execution.aggregate.BaseAggregateExec
+import org.apache.spark.sql.execution.streaming.{MemoryStream, SessionWindowStateStoreRestoreExec, SessionWindowStateStoreSaveExec}
+import org.apache.spark.sql.functions.{count, session_window}
+import org.apache.spark.sql.internal.SQLConf
+import org.apache.spark.sql.streaming.util.StatefulOpClusteredDistributionTestHelper
+import org.apache.spark.util.Utils
+
+class StreamingSessionWindowDistributionSuite extends StreamTest
+ with StatefulOpClusteredDistributionTestHelper with Logging {
+
+ import testImplicits._
+
+ test("SPARK-38204: session window aggregation should require StatefulOpClusteredDistribution " +
+ "from children") {
+
+ withSQLConf(
+ // exclude partial merging session to simplify test
+ SQLConf.STREAMING_SESSION_WINDOW_MERGE_SESSIONS_IN_LOCAL_PARTITION.key -> "false") {
+
+ val inputData = MemoryStream[(String, String, Long)]
+
+ // Split the lines into words, treat words as sessionId of events
+ val events = inputData.toDF()
+ .select($"_1".as("value"), $"_2".as("userId"), $"_3".as("timestamp"))
+ .withColumn("eventTime", $"timestamp".cast("timestamp"))
+ .withWatermark("eventTime", "30 seconds")
+ .selectExpr("explode(split(value, ' ')) AS sessionId", "userId", "eventTime")
+
+ val sessionUpdates = events
+ .repartition($"userId")
+ .groupBy(session_window($"eventTime", "10 seconds") as 'session, 'sessionId, 'userId)
+ .agg(count("*").as("numEvents"))
+ .selectExpr("sessionId", "userId", "CAST(session.start AS LONG)",
+ "CAST(session.end AS LONG)",
+ "CAST(session.end AS LONG) - CAST(session.start AS LONG) AS durationMs",
+ "numEvents")
+
+ testStream(sessionUpdates, OutputMode.Append())(
+ AddData(inputData,
+ ("hello world spark streaming", "key1", 40L),
+ ("world hello structured streaming", "key2", 41L)
+ ),
+
+ // skip checking the result, since we focus to verify the physical plan
+ ProcessAllAvailable(),
+ Execute { query =>
+ val numPartitions = query.lastExecution.numStateStores
+
+ val operators = query.lastExecution.executedPlan.collect {
+ case s: SessionWindowStateStoreRestoreExec => s
+ case s: SessionWindowStateStoreSaveExec => s
+ }
+
+ assert(operators.nonEmpty)
+ operators.foreach { stateOp =>
+ assert(requireStatefulOpClusteredDistribution(stateOp, Seq(Seq("sessionId", "userId")),
+ numPartitions))
+ assert(hasDesiredHashPartitioningInChildren(stateOp, Seq(Seq("sessionId", "userId")),
+ numPartitions))
+ }
+
+ // Verify aggregations in between, except partial aggregation.
+ // This includes MergingSessionsExec.
+ val allAggregateExecs = query.lastExecution.executedPlan.collect {
+ case a: BaseAggregateExec => a
+ }
+
+ val aggregateExecsWithoutPartialAgg = allAggregateExecs.filter {
+ _.requiredChildDistribution.head != UnspecifiedDistribution
+ }
+
+ // We expect single partial aggregation since we disable partial merging sessions.
+ // Remaining agg execs should have child producing expected output partitioning.
+ assert(allAggregateExecs.length - 1 === aggregateExecsWithoutPartialAgg.length)
+
+ // For aggregate execs, we make sure output partitioning of the children is same as
+ // we expect, HashPartitioning with clustering keys & number of partitions.
+ aggregateExecsWithoutPartialAgg.foreach { aggr =>
+ assert(hasDesiredHashPartitioningInChildren(aggr, Seq(Seq("sessionId", "userId")),
+ numPartitions))
+ }
+ }
+ )
+ }
+ }
+
+ test("SPARK-38204: session window aggregation should require ClusteredDistribution " +
+ "from children if the query starts from checkpoint in 3.2") {
+
+ withSQLConf(
+ // exclude partial merging session to simplify test
+ SQLConf.STREAMING_SESSION_WINDOW_MERGE_SESSIONS_IN_LOCAL_PARTITION.key -> "false",
+ // explode(split(value, ' ')) with value as nullable field seems to give different
+ // nullability between Spark 3.2 and 3.3 (nullable in 3.2 and non-null in 3.3)
+ // This should be allowed, but unfortunately, nullability check in state store schema
+ // checker seems to have a bug (`from` and `to` are swapped). Applying temporary
+ // workaround for now.
+ // TODO: investigate the schema checker and remove this once it is resolved.
Review comment:
SPARK-38412 - https://github.com/apache/spark/pull/35731
--
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