Github user harishreedharan commented on a diff in the pull request:
https://github.com/apache/spark/pull/9143#discussion_r43964535
--- Diff:
streaming/src/main/scala/org/apache/spark/streaming/util/BatchedWriteAheadLog.scala
---
@@ -0,0 +1,212 @@
+/*
+ * 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.streaming.util
+
+import java.nio.ByteBuffer
+import java.util.concurrent.{LinkedBlockingQueue, TimeoutException}
+import java.util.{Iterator => JIterator}
+
+import scala.collection.JavaConverters._
+import scala.collection.mutable.ArrayBuffer
+import scala.concurrent.{Await, ExecutionContext, Promise}
+import scala.concurrent.duration._
+import scala.util.control.NonFatal
+
+import org.apache.spark.Logging
+import org.apache.spark.util.{Utils, ThreadUtils}
+
+/**
+ * A wrapper for a WriteAheadLog that batches records before writing data.
All other methods will
+ * be passed on to the wrapped class.
+ *
+ * Parent exposed for testing.
+ */
+private[streaming] class BatchedWriteAheadLog(private[util] val parent:
WriteAheadLog)
+ extends WriteAheadLog with Logging {
+
+ import BatchedWriteAheadLog._
+
+ // exposed for tests
+ protected val walWriteQueue = new LinkedBlockingQueue[RecordBuffer]()
+
+ private val WAL_WRITE_STATUS_TIMEOUT = 5000 // 5 seconds
+
+ // Whether the writer thread is active
+ @volatile private var active: Boolean = true
+ protected val buffer = new ArrayBuffer[RecordBuffer]()
+
+ private val batchedWriterThread = startBatchedWriterThread()
+
+ /**
+ * Write a byte buffer to the log file. This method adds the byteBuffer
to a queue and blocks
+ * until the record is properly written by the parent.
+ */
+ override def write(byteBuffer: ByteBuffer, time: Long):
WriteAheadLogRecordHandle = {
+ val promise = Promise[WriteAheadLogRecordHandle]()
+ walWriteQueue.offer(RecordBuffer(byteBuffer, time, promise))
+ try {
+ Await.result(promise.future.recover { case _ => null
}(ThreadUtils.sameThread),
+ WAL_WRITE_STATUS_TIMEOUT.milliseconds)
+ } catch {
+ case e: TimeoutException =>
+ logWarning(s"Write to Write Ahead Log promise timed out after " +
+ s"$WAL_WRITE_STATUS_TIMEOUT millis for record.")
+ null
+ }
+ }
+
+ /**
+ * Read a segment from an existing Write Ahead Log. The data may be
aggregated, and the user
+ * should de-aggregate using [[BatchedWriteAheadLog.deaggregate]]
+ *
+ * This method is handled by the parent WriteAheadLog.
+ */
+ override def read(segment: WriteAheadLogRecordHandle): ByteBuffer = {
+ parent.read(segment)
+ }
+
+ /**
+ * Read all the existing logs from the log directory.
+ *
+ * This method is handled by the parent WriteAheadLog.
+ */
+ override def readAll(): JIterator[ByteBuffer] = {
+ parent.readAll().asScala.flatMap(deaggregate).asJava
--- End diff --
I really don't think we are dealing with all that much memory. I feel we
are over-using the conversions here with very little additional benefit. All we
are doing is a flatMap.
Moreover, the flatMap is going to allocate a new collection/seq - so a loop
is probably going to use less memory, because you are adding only the newly
generated `ByteBuffers`. There are no new objects allocated other than the ones
required (in the last comment when I said records, I meant the Buffers that
were created as a result of the deaggregation, not all of the data from the
`readAll`)
---
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 [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
