turboFei commented on a change in pull request #25863:
[SPARK-28945][SPARK-29037][CORE][SQL] Fix the issue that spark gives duplicate
result and support concurrent file source write operations write to different
partitions in the same table.
URL: https://github.com/apache/spark/pull/25863#discussion_r332397399
##########
File path:
sql/core/src/main/scala/org/apache/spark/sql/execution/datasources/InsertIntoHadoopFsRelationCommand.scala
##########
@@ -266,4 +304,131 @@ case class InsertIntoHadoopFsRelationCommand(
}
}.toMap
}
+
+ /**
+ * Check current committer whether supports several
InsertIntoHadoopFsRelation operations write
+ * to different partitions in a same table concurrently. If supports, then
detect the conflict
+ * whether there are several operations write to same partition in the same
table or write to
+ * a non-partitioned table.
+ */
+ private def detectConflict(
+ commitProtocol: FileCommitProtocol,
+ fs: FileSystem,
+ path: Path,
+ staticPartitionKVs: Seq[(String, String)],
+ appId: String,
+ jobId: String): Unit = {
+ import HadoopMapReduceCommitProtocol._
+
+ val supportConcurrent =
commitProtocol.isInstanceOf[HadoopMapReduceCommitProtocol] &&
+
commitProtocol.asInstanceOf[HadoopMapReduceCommitProtocol].supportConcurrent
+ if (supportConcurrent && fs.exists(outputPath) &&
fs.isDirectory(outputPath)) {
+
+ val insertStagingPath = ".spark-staging-" + staticPartitionKVs.size
+ val checkedPath = new Path(outputPath, Array(insertStagingPath,
+
getEscapedStaticPartitionPath(staticPartitionKVs)).mkString(File.separator))
+ insertStagingDir = new Path(outputPath, insertStagingPath)
+
+ if (fs.exists(checkedPath)) {
+ throwConflictException(fs, insertStagingDir, staticPartitionKVs.size,
+ staticPartitionKVs)
+ }
+ stagingOutputDir = new Path(outputPath, Array(insertStagingPath,
+ getEscapedStaticPartitionPath(staticPartitionKVs), appId,
jobId).mkString(File.separator))
+ fs.mkdirs(stagingOutputDir)
+
+ for (i <- 0 to partitionColumns.size) {
+ if (i != staticPartitionKVs.size) {
+ val stagingDir = new Path(path, ".spark-staging-" + i)
+ if (fs.exists(stagingDir)) {
+ val subPath = getEscapedStaticPartitionPath(
+ staticPartitionKVs.slice(0, i))
+ val checkedPath = if (!subPath.isEmpty) {
+ new Path(stagingDir, subPath)
+ } else {
+ stagingDir
+ }
+ if (fs.exists(checkedPath)) {
+ throwConflictException(fs, stagingDir, i, staticPartitionKVs)
+ }
+ }
+ }
+ }
+ }
+ }
+
+ private def throwConflictException(
+ fs: FileSystem,
+ stagingDir: Path,
+ depth: Int,
+ staticPartitionKVs: Seq[(String, String)]): Unit = {
+ val conflictedPaths = ListBuffer[Path]()
+ val currentPath = if (depth == staticPartitionKVs.size ||
staticPartitionKVs.size == 0) {
+ stagingDir
+ } else {
+ new Path(stagingDir,
HadoopMapReduceCommitProtocol.getEscapedStaticPartitionPath(
+ staticPartitionKVs.slice(0, staticPartitionKVs.size - depth)))
+ }
+
+ findConflictedStagingOutputPaths(fs, currentPath, depth, conflictedPaths)
+
+ val pathsInfo = conflictedPaths.toList
+ .map { path =>
+ val absolutePath = path.toUri.getRawPath
+ val relativePath =
absolutePath.substring(absolutePath.lastIndexOf(stagingDir.getName))
+ var appId: Option[String] = None
+ var modificationTime: Date = null
+ try {
+ val files = fs.listStatus(path)
+ if (files.size > 0) {
+ appId = Some(files.apply(0).getPath.getName)
+ }
+ modificationTime = new
Date(fs.getFileStatus(path).getModificationTime)
+ } catch {
+ case e: Exception => logWarning("Exception occurred", e)
+ }
+ (relativePath, appId.getOrElse("Not Found"), modificationTime)
+ }
+
+ throw new InsertFileSourceConflictException(
+ s"""
+ | Conflict is detected, some other conflicted output path(s) under
tablePath:
+ | ($outputPath) existed.
+ | Relative path, appId and last modification time information is
shown as below:
+ | ${pathsInfo}.
+ | There may be two possibilities:
+ | 1. Another InsertDataSource operation is executing, you need wait
for it to
+ | complete.
+ | 2. This dir is belong to a killed application and not be cleaned up
gracefully.
+ |
+ | Please check the last modification time and use given appId to
judge whether
+ | relative application is running now. If not, you should delete
responding path
+ | without recursive manually.
+ |""".stripMargin)
+ }
+
+ /**
+ * Find relative staging output paths, which is conflicted with current
+ * InsertIntoHadoopFsRelation operation.
+ */
+ private def findConflictedStagingOutputPaths(
+ fs: FileSystem,
+ path: Path,
+ depth: Int,
+ paths: ListBuffer[Path]): Unit = {
+ try {
+ if (fs.exists(path)) {
+ if (depth == 0) {
+ paths += path
+ } else {
+ for (file <- fs.listStatus(path)) {
Review comment:
I checked whether this path is existed firstly and caught all exceptions in
this code block.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
