Github user huafengw commented on a diff in the pull request:
https://github.com/apache/incubator-gearpump/pull/190#discussion_r123429442
--- Diff:
streaming/src/main/scala/org/apache/gearpump/streaming/refactor/state/StatefulTask.scala
---
@@ -0,0 +1,179 @@
+/*
+ * 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.gearpump.streaming.refactor.state
+
+import java.io._
+import java.time.Instant
+import java.util
+import java.util.Map
+
+import com.google.common.collect.Table
+import org.apache.gearpump.cluster.UserConfig
+import org.apache.gearpump.streaming.refactor.coder._
+import org.apache.gearpump.streaming.refactor.state.api.StateInternals
+import
org.apache.gearpump.streaming.refactor.state.heap.HeapStateInternalsFactory
+import org.apache.gearpump.streaming.state.impl.{CheckpointManager,
PersistentStateConfig}
+import org.apache.gearpump.streaming.task.{Task, TaskContext,
UpdateCheckpointClock}
+import org.apache.gearpump.streaming.transaction.api.CheckpointStoreFactory
+import org.apache.gearpump.util.LogUtil
+import org.apache.gearpump.{Message, TimeStamp}
+
+object StatefulTask {
+ val LOG = LogUtil.getLogger(getClass)
+}
+
+/**
+ * stateful task that support state access and all state will be backed in
memory
+ * after checkpoint state will persist into storage layer and it will
guarantee
+ * 'exactly-once' process semantic
+ *
+ */
+abstract class StatefulTask(taskContext: TaskContext, conf: UserConfig)
+ extends Task(taskContext, conf) {
+
+ import taskContext._
+
+ val checkpointStoreFactory = conf.getValue[CheckpointStoreFactory](
+ PersistentStateConfig.STATE_CHECKPOINT_STORE_FACTORY).get
+ val checkpointStore = checkpointStoreFactory.getCheckpointStore(
+ s"app$appId-task${taskId.processorId}_${taskId.index}")
+ val checkpointInterval =
conf.getLong(PersistentStateConfig.STATE_CHECKPOINT_INTERVAL_MS).get
+ val checkpointManager = new CheckpointManager(checkpointInterval,
checkpointStore)
+
+ var inited = false
+
+ // core state data
+ var encodedKeyStateMap: Map[String, Table[String, String, Array[Byte]]]
= null
+
+ /**
+ * subclass should override this method if they want to init state api
+ */
+ def open: Unit = {}
+
+ /**
+ * Subclass should override this method to specify how a new message
should update state
+ */
+ def invoke(message: Message): Unit
+
+ def close: Unit = {}
+
+ def getStateInternals[KT](keyCoder: Coder[KT], key: KT): StateInternals
= {
+ if (!inited) {
+ throw new RuntimeException(" please init state access object in
`open` method! ")
+ }
+ if (encodedKeyStateMap == null) {
+ encodedKeyStateMap = new util.HashMap[String, Table[String, String,
Array[Byte]]]()
+ }
+
+ val factory = new HeapStateInternalsFactory[KT](keyCoder,
encodedKeyStateMap)
+ factory.stateInternalsForKey(key)
+ }
+
+ final override def onStart(startTime: Instant): Unit = {
+ // recover state from snapshot
+ LOG.info("[onStart] - recover from snapshot")
+ val timestamp = startTime.toEpochMilli
+ checkpointManager
+ .recover(timestamp)
+ .foreach(recoverState(timestamp, _))
+ reportCheckpointClock(timestamp)
+ inited = true
+ open
+ }
+
+ final override def onNext(message: Message): Unit = {
+ checkpointManager.update(message.timestamp.toEpochMilli)
+ invoke(message)
+ }
+
+ final override def onWatermarkProgress(watermark: Instant): Unit = {
+ if (checkpointManager.shouldCheckpoint(watermark.toEpochMilli)) {
+ checkpointManager.getCheckpointTime.foreach { checkpointTime =>
+ val serialized = snapshot
+ checkpointManager.checkpoint(checkpointTime, serialized)
+ reportCheckpointClock(checkpointTime)
+ }
+ }
+ }
+
+ final override def onStop(): Unit = {
+ LOG.info("[onStop] closing checkpoint manager")
+ close
+ checkpointManager.close()
+ }
+
+ private def reportCheckpointClock(timestamp: TimeStamp): Unit = {
+ LOG.debug("reportCheckpointClock at : {}", timestamp)
+ appMaster ! UpdateCheckpointClock(taskContext.taskId, timestamp)
+ }
+
+ private def recoverState(timestamp: TimeStamp, snapshot: Array[Byte]):
Unit = {
+ LOG.info("call recoverState snapshot with timestamp : {}", timestamp)
+ var bis: Option[ByteArrayInputStream] = None
+ var oin: Option[ObjectInputStream] = None
+ try {
+ bis = Some(new ByteArrayInputStream(snapshot))
+ oin = Some(new ObjectInputStream(bis.get))
+ encodedKeyStateMap = oin.get.readObject().asInstanceOf[
+ Map[String, Table[String, String, Array[Byte]]]]
+ } catch {
+ case ex: IOException => throw new RuntimeException(ex)
+ case e: Exception => throw new RuntimeException(e)
+ } finally {
+ try {
+ if (oin.nonEmpty) oin.get.close()
+ } catch {
+ case _ =>
+ }
+ try {
+ if (bis.nonEmpty) bis.get.close()
+ } catch {
+ case _ =>
+ }
+ }
+ }
+
+ private def snapshot: Array[Byte] = {
+ LOG.info("do snapshot ")
+ var buffer: Option[ByteArrayOutputStream] = None
+ var oout: Option[ObjectOutputStream] = None
+ try {
+ buffer = Some(new ByteArrayOutputStream)
+ oout = Some(new ObjectOutputStream(buffer.get))
+ oout.get.writeObject(encodedKeyStateMap)
+ oout.get.flush()
+ buffer.get.toByteArray
+ } catch {
+ case ex: IOException => throw new RuntimeException(ex)
+ case e: Exception => throw new RuntimeException(e)
+ } finally {
+ try {
+ if (oout.nonEmpty) oout.get.close()
--- End diff --
The same
---
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.
---
