[ https://issues.apache.org/jira/browse/SPARK-26327?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16721785#comment-16721785 ]
ASF GitHub Bot commented on SPARK-26327: ---------------------------------------- dongjoon-hyun closed pull request #23299: [SPARK-26327][SQL][BACKPORT-2.3] Bug fix for `FileSourceScanExec` metrics update URL: https://github.com/apache/spark/pull/23299 This is a PR merged from a forked repository. As GitHub hides the original diff on merge, it is displayed below for the sake of provenance: As this is a foreign pull request (from a fork), the diff is supplied below (as it won't show otherwise due to GitHub magic): diff --git a/sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala b/sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala index 0a23264a4a16b..5543087db0614 100644 --- a/sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala +++ b/sql/core/src/main/scala/org/apache/spark/sql/execution/DataSourceScanExec.scala @@ -183,19 +183,14 @@ case class FileSourceScanExec( partitionSchema = relation.partitionSchema, relation.sparkSession.sessionState.conf) + private var metadataTime = 0L + @transient private lazy val selectedPartitions: Seq[PartitionDirectory] = { val optimizerMetadataTimeNs = relation.location.metadataOpsTimeNs.getOrElse(0L) val startTime = System.nanoTime() val ret = relation.location.listFiles(partitionFilters, dataFilters) val timeTakenMs = ((System.nanoTime() - startTime) + optimizerMetadataTimeNs) / 1000 / 1000 - - metrics("numFiles").add(ret.map(_.files.size.toLong).sum) - metrics("metadataTime").add(timeTakenMs) - - val executionId = sparkContext.getLocalProperty(SQLExecution.EXECUTION_ID_KEY) - SQLMetrics.postDriverMetricUpdates(sparkContext, executionId, - metrics("numFiles") :: metrics("metadataTime") :: Nil) - + metadataTime = timeTakenMs ret } @@ -293,6 +288,8 @@ case class FileSourceScanExec( } private lazy val inputRDD: RDD[InternalRow] = { + // Update metrics for taking effect in both code generation node and normal node. + updateDriverMetrics() val readFile: (PartitionedFile) => Iterator[InternalRow] = relation.fileFormat.buildReaderWithPartitionValues( sparkSession = relation.sparkSession, @@ -500,6 +497,19 @@ case class FileSourceScanExec( } } + /** + * Send the updated metrics to driver, while this function calling, selectedPartitions has + * been initialized. See SPARK-26327 for more detail. + */ + private def updateDriverMetrics() = { + metrics("numFiles").add(selectedPartitions.map(_.files.size.toLong).sum) + metrics("metadataTime").add(metadataTime) + + val executionId = sparkContext.getLocalProperty(SQLExecution.EXECUTION_ID_KEY) + SQLMetrics.postDriverMetricUpdates(sparkContext, executionId, + metrics("numFiles") :: metrics("metadataTime") :: Nil) + } + override def doCanonicalize(): FileSourceScanExec = { FileSourceScanExec( relation, diff --git a/sql/core/src/test/scala/org/apache/spark/sql/execution/metric/SQLMetricsSuite.scala b/sql/core/src/test/scala/org/apache/spark/sql/execution/metric/SQLMetricsSuite.scala index a3a3f3851e21c..439a36080ce90 100644 --- a/sql/core/src/test/scala/org/apache/spark/sql/execution/metric/SQLMetricsSuite.scala +++ b/sql/core/src/test/scala/org/apache/spark/sql/execution/metric/SQLMetricsSuite.scala @@ -504,4 +504,19 @@ class SQLMetricsSuite extends SparkFunSuite with SQLMetricsTestUtils with Shared test("writing data out metrics with dynamic partition: parquet") { testMetricsDynamicPartition("parquet", "parquet", "t1") } + + test("SPARK-26327: FileSourceScanExec metrics") { + withTable("testDataForScan") { + spark.range(10).selectExpr("id", "id % 3 as p") + .write.partitionBy("p").saveAsTable("testDataForScan") + // The execution plan only has 1 FileScan node. + val df = spark.sql( + "SELECT * FROM testDataForScan WHERE p = 1") + testSparkPlanMetrics(df, 1, Map( + 0L -> (("Scan parquet default.testdataforscan", Map( + "number of output rows" -> 3L, + "number of files" -> 2L)))) + ) + } + } } ---------------------------------------------------------------- This is an automated message from the Apache Git Service. To respond to the message, please log on GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org > Metrics in FileSourceScanExec not update correctly while > relation.partitionSchema is set > ---------------------------------------------------------------------------------------- > > Key: SPARK-26327 > URL: https://issues.apache.org/jira/browse/SPARK-26327 > Project: Spark > Issue Type: Sub-task > Components: SQL > Affects Versions: 2.4.0 > Reporter: Yuanjian Li > Assignee: Yuanjian Li > Priority: Major > Fix For: 3.0.0 > > > As currently approach in `FileSourceScanExec`, the metrics of "numFiles" and > "metadataTime"(fileListingTime) were updated while lazy val > `selectedPartitions` initialized in the scenario of relation.partitionSchema > is set. But `selectedPartitions` will be initialized by `metadata` at first, > which is called by `queryExecution.toString` in > `SQLExecution.withNewExecutionId`. So while the > `SQLMetrics.postDriverMetricUpdates` called, there's no corresponding > liveExecutions in SQLAppStatusListener, the metrics update is not work. -- This message was sent by Atlassian JIRA (v7.6.3#76005) --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org For additional commands, e-mail: issues-h...@spark.apache.org