Mailing-List: contact reviews-help@spark.apache.org; run by ezmlm
Precedence: bulk
Reply-To: reviews@spark.apache.org
From: vanzin <git@git.apache.org>
To: reviews@spark.apache.org
Reply-To: reviews@spark.apache.org
References: <git-pr-718-spark@git.apache.org>
In-Reply-To: <git-pr-718-spark@git.apache.org>
Subject: [GitHub] spark pull request: [SPARK-1768] History server
 enhancements.
Content-Type: text/plain
Message-Id: <20140610195921.6457C91DC19@tyr.zones.apache.org>
Date: Tue, 10 Jun 2014 19:59:21 +0000 (UTC)

Github user vanzin commented on a diff in the pull request:

    https://github.com/apache/spark/pull/718#discussion_r13615472
  
    --- Diff: core/src/main/scala/org/apache/spark/deploy/history/FsHistoryProvider.scala ---
    @@ -0,0 +1,193 @@
    +/*
    + * 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.FileNotFoundException
    +
    +import scala.collection.mutable
    +
    +import org.apache.hadoop.fs.{FileStatus, Path}
    +
    +import org.apache.spark.{Logging, SecurityManager, SparkConf}
    +import org.apache.spark.scheduler._
    +import org.apache.spark.ui.SparkUI
    +import org.apache.spark.util.Utils
    +
    +private[history] class FsHistoryProvider(conf: SparkConf) extends ApplicationHistoryProvider
    +  with Logging {
    +
    +  // Interval between each check for event log updates
    +  private val UPDATE_INTERVAL_MS = conf.getInt("spark.history.fs.updateInterval", 10) * 1000
    +
    +  private val logDir = conf.get("spark.history.fs.logDirectory")
    +  private val fs = Utils.getHadoopFileSystem(logDir)
    +
    +  // A timestamp of when the disk was last accessed to check for log updates
    +  private var lastLogCheckTimeMs = -1L
    +
    +  // List of applications, in order from newest to oldest.
    +  @volatile private var appList: Seq[ApplicationHistoryInfo] = Nil
    +
    +  /**
    +   * A background thread that periodically checks for event log updates on disk.
    +   *
    +   * If a log check is invoked manually in the middle of a period, this thread re-adjusts the
    +   * time at which it performs the next log check to maintain the same period as before.
    +   *
    +   * TODO: Add a mechanism to update manually.
    +   */
    +  private val logCheckingThread = new Thread("LogCheckingThread") {
    +    override def run() = Utils.logUncaughtExceptions {
    +      while (true) {
    +        val now = getMonotonicTime()
    +        if (now - lastLogCheckTimeMs > UPDATE_INTERVAL_MS) {
    +          Thread.sleep(UPDATE_INTERVAL_MS)
    +        } else {
    +          // If the user has manually checked for logs recently, wait until
    +          // UPDATE_INTERVAL_MS after the last check time
    +          Thread.sleep(lastLogCheckTimeMs + UPDATE_INTERVAL_MS - now)
    +        }
    +        checkForLogs()
    +      }
    +    }
    +  }
    +
    +  initialize()
    +
    +  private def initialize() {
    +    // Validate the log directory.
    +    val path = new Path(logDir)
    +    if (!fs.exists(path)) {
    +      throw new IllegalArgumentException(
    +        "Logging directory specified does not exist: %s".format(logDir))
    +    }
    +    if (!fs.getFileStatus(path).isDir) {
    +      throw new IllegalArgumentException(
    +        "Logging directory specified is not a directory: %s".format(logDir))
    +    }
    +
    +    checkForLogs()
    +    logCheckingThread.setDaemon(true)
    +    logCheckingThread.start()
    +  }
    +
    +  override def getListing() = appList
    +
    +  override def getAppInfo(appId: String): ApplicationHistoryInfo = {
    +    try {
    +      val appLogDir = fs.getFileStatus(new Path(logDir, appId))
    +      loadAppInfo(appLogDir, true)
    +    } catch {
    +      case e: FileNotFoundException => null
    +    }
    +  }
    +
    +  /**
    +   * Builds the application list based on the current contents of the log directory.
    +   * Tries to reuse as much of the data already in memory as possible, by not reading
    +   * applications that haven't been updated since last time the logs were checked.
    +   */
    +  def checkForLogs() = {
    +    lastLogCheckTimeMs = getMonotonicTime()
    +    logDebug("Checking for logs. Time is now %d.".format(lastLogCheckTimeMs))
    +    try {
    +      val logStatus = fs.listStatus(new Path(logDir))
    +      val logDirs = if (logStatus != null) logStatus.filter(_.isDir).toSeq else Seq[FileStatus]()
    +      val logInfos = logDirs
    +        .filter {
    +            dir => fs.isFile(new Path(dir.getPath(), EventLoggingListener.APPLICATION_COMPLETE))
    +          }
    +
    +      val currentApps = Map[String, ApplicationHistoryInfo](
    +        appList.map(app => (app.id -> app)):_*)
    --- End diff --
    
    I don't think there's a good solution here without a bigger change. The issue is that we want to list apps in some sensible order (current code uses end time), but when checking whether the app is up-to-date we need to compare its id. And just using a map doesn't allow us to have both.
    
    So to avoid using more memory we'd have to reload every app from HDFS when rescanning, which can be pretty slow (and, with the check interval being 10s by default, pretty unfriendly to the NN).
    
    At this point I'd like to defer a solution to this particular problem to a separate change. I've been playing with some ideas to solve this and other issues in the code, but that part is not yet ready for review.


---
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 infrastructure@apache.org or file a JIRA ticket
with INFRA.
---