HeartSaVioR commented on a change in pull request #26416: [WIP][SPARK-29779][CORE] Compact old event log files and cleanup URL: https://github.com/apache/spark/pull/26416#discussion_r345523179
########## File path: core/src/main/scala/org/apache/spark/deploy/history/EventLogFileCompactor.scala ########## @@ -0,0 +1,235 @@ +/* + * 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.deploy.history + +import java.io.IOException +import java.net.URI +import java.util.ServiceLoader + +import scala.collection.JavaConverters._ +import scala.io.{Codec, Source} + +import com.fasterxml.jackson.databind.exc.UnrecognizedPropertyException +import org.apache.hadoop.conf.Configuration +import org.apache.hadoop.fs.{FileStatus, FileSystem, Path} +import org.json4s.jackson.JsonMethods.parse + +import org.apache.spark.SparkConf +import org.apache.spark.internal.Logging +import org.apache.spark.internal.config.EVENT_LOG_ROLLING_MAX_FILES_TO_RETAIN +import org.apache.spark.scheduler._ +import org.apache.spark.util.{JsonProtocol, Utils} + +class EventLogFileCompactor( + sparkConf: SparkConf, + hadoopConf: Configuration, + fs: FileSystem) extends Logging { + + private val maxFilesToRetain: Int = sparkConf.get(EVENT_LOG_ROLLING_MAX_FILES_TO_RETAIN) + + // FIXME: javadoc - caller should provide event log files (either compacted or original) + // sequentially if the last event log file is already a compacted file, everything + // will be skipped + def compact(eventLogFiles: Seq[FileStatus]): Seq[FileStatus] = { + if (eventLogFiles.isEmpty) { + return Seq.empty[FileStatus] + } + + // skip everything if the last file is already a compacted file + if (EventLogFileWriter.isCompacted(eventLogFiles.last.getPath)) { + return Seq(eventLogFiles.last) + } + + val (filesToCompact, filesToRetain) = findFilesToCompact(eventLogFiles) + if (filesToCompact.isEmpty) { + filesToRetain + } else { + // first pass + val bus = new ReplayListenerBus() + + val builders = ServiceLoader.load(classOf[EventFilterBuilder], + Utils.getContextOrSparkClassLoader).asScala + builders.foreach(bus.addListener) + + filesToCompact.foreach { log => + Utils.tryWithResource(EventLogFileReader.openEventLog(log.getPath, fs)) { in => + bus.replay(in, log.getPath.getName) + } + } + + // second pass + val rewriter = new FilteredEventLogFileRewriter(sparkConf, hadoopConf, + filesToCompact, fs, builders.map(_.createFilter()).toSeq) + rewriter.start() + rewriter.rewrite() + rewriter.stop() + + // cleanup files which are replaced with new compacted file. + cleanupCompactedFiles(filesToCompact) + + fs.getFileStatus(new Path(rewriter.logPath)) :: filesToRetain.toList + } + } + + private def cleanupCompactedFiles(files: Seq[FileStatus]): Unit = { + files.foreach { file => + try { + fs.delete(file.getPath, true) + } catch { + case _: IOException => logWarning(s"Failed to remove ${file.getPath} / skip removing.") + } + } + } + + private def findFilesToCompact( + eventLogFiles: Seq[FileStatus]): (Seq[FileStatus], Seq[FileStatus]) = { + val lastCompactedFileIdx = eventLogFiles.lastIndexWhere { fs => + EventLogFileWriter.isCompacted(fs.getPath) + } + val files = eventLogFiles.drop(lastCompactedFileIdx) + + if (files.length > maxFilesToRetain) { + (files.dropRight(maxFilesToRetain), files.takeRight(maxFilesToRetain)) Review comment: You cannot cleanup log file unless compacted. So as long as `maxFilesToRetain` has the semantic of cleaning old files, that is the minimum, so in fact it's opposite. We may be able to add one more config to let compactor be able to compact over retained files to speed up replaying, though I think the original approach would be better in point of replaying the logs. Once I finish this I would try to go back and see how we can do incremental replay with snapshot approach. ---------------------------------------------------------------- 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]
