Github user tdas commented on a diff in the pull request:
https://github.com/apache/spark/pull/11645#discussion_r56227475
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/streaming/state/HDFSBackedStateStoreProvider.scala
---
@@ -0,0 +1,454 @@
+/*
+ * 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.streaming.state
+
+import scala.collection.mutable
+import scala.util.Random
+import scala.util.control.NonFatal
+
+import org.apache.hadoop.conf.Configuration
+import org.apache.hadoop.fs.{FileStatus, Path}
+
+import org.apache.spark.{Logging, SparkConf}
+import org.apache.spark.serializer.{DeserializationStream, KryoSerializer,
SerializationStream}
+import org.apache.spark.sql.catalyst.InternalRow
+import org.apache.spark.sql.catalyst.expressions.JoinedRow
+import org.apache.spark.util.{CompletionIterator, Utils}
+
+
+/**
+ * An implementation of [[StateStoreProvider]] and [[StateStore]] in which
all the data is backed
+ * by files in a HDFS-compatible file system. All updates to the store
has to be done in sets
+ * transactionally, and each set of updates increments the store's
version. These versions can
+ * be used to re-execute the updates (by retries in RDD operations) on the
correct version of
+ * the store, and regenerate the store version.
+ *
+ * Usage:
+ * To update the data in the state store, the following order of
operations are needed.
+ *
+ * - val store = StateStore.get(operatorId, partitionId, version) // to
get the right store
+ * - store.update(...)
+ * - store.remove(...)
+ * - store.commit() // commits all the updates to made with version
number
+ * - store.iterator() // key-value data after last commit as an iterator
+ * - store.updates() // updates made in the last as an iterator
+ *
+ * Fault-tolerance model:
+ * - Every set of updates is written to a delta file before committing.
+ * - The state store is responsible for managing, collapsing and cleaning
up of delta files.
+ * - Multiple attempts to commit the same version of updates must have the
same updates.
+ * - Background management of files ensures that last versions of the
store is always recoverable
+ * to ensure re-executed RDD operations re-apply updates on the correct
past version of the
+ * store.
+ */
+class HDFSBackedStateStoreProvider(
+ val id: StateStoreId,
+ val directory: String,
+ numBatchesToRetain: Int = 2,
+ maxDeltaChainForSnapshots: Int = 10
+ ) extends StateStoreProvider with Logging {
+ type MapType = mutable.HashMap[InternalRow, InternalRow]
+
+ import StateStore._
+
+
+ class HDFSBackedStateStore( val version: Long, mapToUpdate: MapType)
+ extends StateStore {
+
+ trait STATE
+ case object UPDATING extends STATE
+ case object COMMITTED extends STATE
+ case object CANCELLED extends STATE
+
+ private val newVersion = version + 1
+ private val tempDeltaFile = new Path(baseDir,
s"temp-${Random.nextLong}")
+ private val tempDeltaFileStream =
+ serializer.newInstance().serializeStream(fs.create(tempDeltaFile,
true))
+ private val allUpdates = new mutable.HashMap[InternalRow, StoreUpdate]
+
+ @volatile private var state: STATE = UPDATING
+ @volatile private var finalDeltaFile: Path = null
+
+ override def id: StateStoreId = HDFSBackedStateStoreProvider.this.id
+
+ /** Update the value of a key using the value generated by the update
function */
+ override def update(key: InternalRow, updateFunc: Option[InternalRow]
=> InternalRow): Unit = {
+ verify(state == UPDATING, "Cannot update after already committed or
cancelled")
+ val oldValueOption = mapToUpdate.get(key)
+ val value = updateFunc(oldValueOption)
+ mapToUpdate.put(key, value)
+ allUpdates.get(key) match {
+ case Some(ValueAdded(_, _)) =>
+ allUpdates.put(key, ValueAdded(key, value))
+ case Some(ValueUpdated(_, _)) | Some(KeyRemoved(_)) =>
+ allUpdates.put(key, ValueUpdated(key, value))
+ case None =>
+ val update =
+ if (oldValueOption.nonEmpty) ValueUpdated(key, value) else
ValueAdded(key, value)
+ allUpdates.put(key, update)
+ }
+ tempDeltaFileStream.writeObject(ValueUpdated(key, value))
+ }
+
+ /** Remove keys that match the following condition */
+ override def remove(condition: InternalRow => Boolean): Unit = {
+ verify(state == UPDATING, "Cannot remove after already committed or
cancelled")
+ val keyIter = mapToUpdate.keysIterator
+ while (keyIter.hasNext) {
+ val key = keyIter.next
+ if (condition(key)) {
+ mapToUpdate.remove(key)
+ allUpdates.get(key) match {
+ case Some(ValueUpdated(_, _)) | None =>
+ allUpdates.put(key, KeyRemoved(key))
+ case Some(ValueAdded(_, _)) =>
+ allUpdates.remove(key)
+ case Some(KeyRemoved(_)) =>
+ // Remove already in update map, no need to change
+ }
+ tempDeltaFileStream.writeObject(KeyRemoved(key))
+ }
+ }
+ }
+
+ /** Commit all the updates that have been made to the store. */
+ override def commit(): Long = {
+ verify(state == UPDATING, "Cannot commit again after already
committed or cancelled")
+
+ try {
+ tempDeltaFileStream.close()
+ finalDeltaFile = commitUpdates(newVersion, mapToUpdate,
tempDeltaFile)
+ state = COMMITTED
+ newVersion
+ } catch {
+ case NonFatal(e) =>
+ throw new IllegalStateException(
+ s"Error committing version $newVersion into
${HDFSBackedStateStoreProvider.this}", e)
+ }
+ }
+
+ /** Cancel all the updates made on this store. This store will not be
usable any more. */
+ override def cancel(): Unit = {
+ state = CANCELLED
+ if (tempDeltaFileStream != null) {
+ tempDeltaFileStream.close()
+ }
+ if (tempDeltaFile != null && fs.exists(tempDeltaFile)) {
+ fs.delete(tempDeltaFile, true)
+ }
+ }
+
+ /**
+ * Get an iterator of all the store data. This can be called only
after committing the
+ * updates.
+ */
+ override def iterator(): Iterator[InternalRow] = {
+ verify(state == COMMITTED, "Cannot get iterator of store data before
comitting")
+ HDFSBackedStateStoreProvider.this.iterator(newVersion)
+ }
+
+ /**
+ * Get an iterator of all the updates made to the store in the current
version.
+ * This can be called only after committing the updates.
+ */
+ override def updates(): Iterator[StoreUpdate] = {
+ verify(state == COMMITTED, "Cannot get iterator of updates before
committing")
+ allUpdates.valuesIterator
+ }
+
+ /**
+ * Whether all updates have been committed
+ */
+ override def hasCommitted: Boolean = {
+ state == COMMITTED
+ }
+ }
+
+ /** Get the state store for making updates to create a new `version` of
the store. */
+ override def getStore(version: Long): StateStore = synchronized {
+ require(version >= 0, "Version cannot be less than 0")
+ val newMap = new MapType()
+ if (version > 0) {
+ newMap ++= loadMap(version)
+ }
+ new HDFSBackedStateStore(version, newMap)
+ }
+
+ override def manage(): Unit = {
+ try {
+ doSnapshot()
+ cleanup()
+ } catch {
+ case NonFatal(e) =>
+ logWarning(s"Error performing snapshot and cleaning up $this")
+ }
+ }
+
+ override def toString(): String = {
+ s"StateStore[id = (op=${id.operatorId},part=${id.partitionId}), dir =
$baseDir]"
+ }
+
+ /* Internal classes and methods */
+
+ private val loadedMaps = new mutable.HashMap[Long, MapType]
+ private val baseDir = new Path(directory,
s"${id.operatorId}/${id.partitionId.toString}")
+ private val fs = baseDir.getFileSystem(new Configuration())
+ private val serializer = new KryoSerializer(new SparkConf)
+
+ initialize()
+
+ private case class StoreFile(version: Long, path: Path, isSnapshot:
Boolean)
+
+ /** Commit a set of updates to the store with the given new version */
+ private def commitUpdates(newVersion: Long, map: MapType, tempDeltaFile:
Path): Path = {
+ synchronized {
+ val finalDeltaFile = deltaFile(newVersion)
+ fs.rename(tempDeltaFile, finalDeltaFile)
+ loadedMaps.put(newVersion, map)
+ finalDeltaFile
+ }
+ }
+
+ /**
+ * Get iterator of all the data of the latest version of the store.
+ * Note that this will look up the files to determined the latest known
version.
+ */
+ private[state] def latestIterator(): Iterator[InternalRow] =
synchronized {
+ val versionsInFiles = fetchFiles().map(_.version).toSet
+ val versionsLoaded = loadedMaps.keySet
+ val allKnownVersions = versionsInFiles ++ versionsLoaded
+ if (allKnownVersions.nonEmpty) {
+ loadMap(allKnownVersions.max)
+ .iterator
+ .map { case (key, value) => new JoinedRow(key, value) }
+ } else Iterator.empty
+ }
+
+ /** Get iterator of a specific version of the store */
+ private[state] def iterator(version: Long): Iterator[InternalRow] =
synchronized {
+ loadMap(version)
+ .iterator
+ .map { case (key, value) => new JoinedRow(key, value) }
+ }
+
+ /** Initialize the store provider */
+ private def initialize(): Unit = {
+ if (!fs.exists(baseDir)) {
+ fs.mkdirs(baseDir)
+ } else {
+ if (!fs.isDirectory(baseDir)) {
+ throw new IllegalStateException(
+ s"Cannot use $directory for storing state data as" +
+ s"$baseDir already exists and is not a directory")
+ }
+ }
+ }
+
+ /** Load the required version of the map data from the backing files */
+ private def loadMap(version: Long): MapType = {
+ if (version <= 0) return new MapType
+ synchronized { loadedMaps.get(version) }.getOrElse {
+ val mapFromFile = readSnapshotFile(version).getOrElse {
+ val prevMap = loadMap(version - 1)
+ val deltaUpdates = readDeltaFile(version)
+ val newMap = new MapType()
+ newMap ++= prevMap
+ newMap.sizeHint(prevMap.size)
+ while (deltaUpdates.hasNext) {
+ deltaUpdates.next match {
+ case ValueAdded(key, value) => newMap.put(key, value)
+ case ValueUpdated(key, value) => newMap.put(key, value)
+ case KeyRemoved(key) => newMap.remove(key)
+ }
+ }
+ newMap
+ }
+ loadedMaps.put(version, mapFromFile)
+ mapFromFile
+ }
+ }
+
+ private def readDeltaFile(version: Long): Iterator[StoreUpdate] = {
+ val fileToRead = deltaFile(version)
+ if (!fs.exists(fileToRead)) {
+ throw new IllegalStateException(
+ s"Cannot read delta file $fileToRead of $this: $fileToRead does
not exist")
+ }
+ val deser = serializer.newInstance()
+ var deserStream: DeserializationStream = null
+ deserStream = deser.deserializeStream(fs.open(fileToRead))
+ val iter = deserStream.asIterator.asInstanceOf[Iterator[StoreUpdate]]
+ CompletionIterator[StoreUpdate, Iterator[StoreUpdate]](
+ iter, {
+ deserStream.close()
+ })
+ }
+
+ private def writeSnapshotFile(version: Long, map: MapType): Unit = {
+ val fileToWrite = snapshotFile(version)
+ val ser = serializer.newInstance()
+ var outputStream: SerializationStream = null
+ Utils.tryWithSafeFinally {
+ outputStream = ser.serializeStream(fs.create(fileToWrite, false))
+ outputStream.writeAll(map.iterator)
+ } {
+ if (outputStream != null) outputStream.close()
+ }
+ }
+
+ private def readSnapshotFile(version: Long): Option[MapType] = {
+ val fileToRead = snapshotFile(version)
+ if (!fs.exists(fileToRead)) return None
+
+ val deser = serializer.newInstance()
+ val map = new MapType()
+ var deserStream: DeserializationStream = null
+
+ try {
+ deserStream = deser.deserializeStream(fs.open(fileToRead))
+ val iter =
deserStream.asIterator.asInstanceOf[Iterator[(InternalRow, InternalRow)]]
+ while (iter.hasNext) {
+ map += iter.next()
+ }
+ Some(map)
+ } finally {
+ if (deserStream != null) deserStream.close()
+ }
+ }
+
+
+ /** Perform a snapshot of the store to allow delta files to be
consolidated */
+ private def doSnapshot(): Unit = {
+ try {
+ val files = fetchFiles()
+ if (files.nonEmpty) {
+ val lastVersion = files.last.version
+ val deltaFilesForLastVersion =
+ filesForVersion(files, lastVersion).filter(_.isSnapshot == false)
+ synchronized { loadedMaps.get(lastVersion) } match {
+ case Some(map) =>
+ if (deltaFilesForLastVersion.size > maxDeltaChainForSnapshots)
{
+ writeSnapshotFile(lastVersion, map)
+ }
+ case None =>
+ // The last map is not loaded, probably some other instance is
incharge
+ }
+
+ }
+ } catch {
+ case NonFatal(e) =>
+ logWarning(s"Error doing snapshots for $this", e)
+ }
+ }
+
+ /**
+ * Clean up old snapshots and delta files that are not needed any more.
It ensures that last
+ * few versions of the store can be recovered from the files, so
re-executed RDD operations
+ * can re-apply updates on the past versions of the store.
+ */
+ private[state] def cleanup(): Unit = {
+ try {
+ val files = fetchFiles()
+ if (files.nonEmpty) {
+ val earliestVersionToRetain = files.last.version -
numBatchesToRetain
+ if (earliestVersionToRetain > 0) {
+ val earliestFileToRetain = filesForVersion(files,
earliestVersionToRetain).head
+ synchronized {
+ loadedMaps.keys.filter(_ <
earliestVersionToRetain).foreach(loadedMaps.remove)
+ }
+ files.filter(_.version < earliestFileToRetain.version).foreach {
f =>
+ fs.delete(f.path, true)
+ }
+ }
+ }
+ } catch {
+ case NonFatal(e) =>
+ logWarning(s"Error cleaning up files for $this", e)
+ }
+ }
+
+ private def filesForVersion(allFiles: Seq[StoreFile], version: Long):
Seq[StoreFile] = {
--- End diff --
add docs
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
