[SPARK-26311][CORE] New feature: apply custom log URL pattern for executor log URLs in SHS

## What changes were proposed in this pull request?

This patch proposes adding a new configuration on SHS: custom executor log URL pattern. This will enable end users to replace executor logs to other than RM provide, like external log service, which enables to serve executor logs when NodeManager becomes unavailable in case of YARN.

End users can build their own of custom executor log URLs with pre-defined patterns which would be vary on each resource manager. This patch adds some patterns to YARN resource manager. (For others, there's even no executor log url available so cannot define patterns as well.)

Please refer the doc change as well as added UTs in this patch to see how to set up the feature.

## How was this patch tested?

Added UT, as well as manual test with YARN cluster

Closes apache#23260 from HeartSaVioR/SPARK-26311.

Authored-by: Jungtaek Lim (HeartSaVioR) <[email protected]>
Signed-off-by: Marcelo Vanzin <[email protected]>

Author: HeartSaVioR

core/src/main/scala/org/apache/spark/SparkContext.scala

@@ -2355,7 +2355,8 @@ class SparkContext(config: SparkConf) extends Logging {
     // Note: this code assumes that the task scheduler has been initialized and has contacted
     // the cluster manager to get an application ID (in case the cluster manager provides one).
     listenerBus.post(SparkListenerApplicationStart(appName, Some(applicationId),
-      startTime, sparkUser, applicationAttemptId, schedulerBackend.getDriverLogUrls))
+      startTime, sparkUser, applicationAttemptId, schedulerBackend.getDriverLogUrls,
+      schedulerBackend.getDriverAttributes))
     _driverLogger.foreach(_.startSync(_hadoopConfiguration))
   }
 

core/src/main/scala/org/apache/spark/deploy/history/FsHistoryProvider.scala

@@ -359,10 +359,9 @@ private[history] class FsHistoryProvider(conf: SparkConf, clock: Clock)
         return None
     }
 
-    val ui = SparkUI.create(None, new AppStatusStore(kvstore), conf, secManager, app.info.name,
-      HistoryServer.getAttemptURI(appId, attempt.info.attemptId),
-      attempt.info.startTime.getTime(),
-      attempt.info.appSparkVersion)
+    val ui = SparkUI.create(None, new HistoryAppStatusStore(conf, kvstore), conf, secManager,
+      app.info.name, HistoryServer.getAttemptURI(appId, attempt.info.attemptId),
+      attempt.info.startTime.getTime(), attempt.info.appSparkVersion)
     loadPlugins().foreach(_.setupUI(ui))
 
     val loadedUI = LoadedAppUI(ui)

core/src/main/scala/org/apache/spark/deploy/history/HistoryAppStatusStore.scala

@@ -0,0 +1,133 @@
+/*
+ * 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.util.concurrent.atomic.AtomicBoolean
+
+import scala.util.matching.Regex
+
+import org.apache.spark.SparkConf
+import org.apache.spark.internal.Logging
+import org.apache.spark.internal.config.History._
+import org.apache.spark.status.AppStatusStore
+import org.apache.spark.status.api.v1
+import org.apache.spark.util.kvstore.KVStore
+
+private[spark] class HistoryAppStatusStore(
+    conf: SparkConf,
+    store: KVStore)
+  extends AppStatusStore(store, None) with Logging {
+
+  import HistoryAppStatusStore._
+
+  private val logUrlPattern: Option[String] = {
+    val appInfo = super.applicationInfo()
+    val applicationCompleted = appInfo.attempts.nonEmpty && appInfo.attempts.head.completed
+    if (applicationCompleted || conf.get(APPLY_CUSTOM_EXECUTOR_LOG_URL_TO_INCOMPLETE_APP)) {
+      conf.get(CUSTOM_EXECUTOR_LOG_URL)
+    } else {
+      None
+    }
+  }
+
+  private val informedForMissingAttributes = new AtomicBoolean(false)
+
+  override def executorList(activeOnly: Boolean): Seq[v1.ExecutorSummary] = {
+    val execList = super.executorList(activeOnly)
+    logUrlPattern match {
+      case Some(pattern) => execList.map(replaceLogUrls(_, pattern))
+      case None => execList
+    }
+  }
+
+  override def executorSummary(executorId: String): v1.ExecutorSummary = {
+    val execSummary = super.executorSummary(executorId)
+    logUrlPattern match {
+      case Some(pattern) => replaceLogUrls(execSummary, pattern)
+      case None => execSummary
+    }
+  }
+
+  private def replaceLogUrls(exec: v1.ExecutorSummary, urlPattern: String): v1.ExecutorSummary = {
+    val attributes = exec.attributes
+
+    // Relation between pattern {{FILE_NAME}} and attribute {{LOG_FILES}}
+    // Given that HistoryAppStatusStore don't know which types of log files can be provided
+    // from resource manager, we require resource manager to provide available types of log
+    // files, which are encouraged to be same as types of log files provided in original log URLs.
+    // Once we get the list of log files, we need to expose them to end users as a pattern
+    // so that end users can compose custom log URL(s) including log file name(s).
+    val allPatterns = CUSTOM_URL_PATTERN_REGEX.findAllMatchIn(urlPattern).map(_.group(1)).toSet
+    val allPatternsExceptFileName = allPatterns.filter(_ != "FILE_NAME")
+    val allAttributeKeys = attributes.keySet
+    val allAttributeKeysExceptLogFiles = allAttributeKeys.filter(_ != "LOG_FILES")
+
+    if (allPatternsExceptFileName.diff(allAttributeKeysExceptLogFiles).nonEmpty) {
+      logFailToRenewLogUrls("some of required attributes are missing in app's event log.",
+        allPatternsExceptFileName, allAttributeKeys)
+      return exec
+    } else if (allPatterns.contains("FILE_NAME") && !allAttributeKeys.contains("LOG_FILES")) {
+      logFailToRenewLogUrls("'FILE_NAME' parameter is provided, but file information is " +
+        "missing in app's event log.", allPatternsExceptFileName, allAttributeKeys)
+      return exec
+    }
+
+    val updatedUrl = allPatternsExceptFileName.foldLeft(urlPattern) { case (orig, patt) =>
+      // we already checked the existence of attribute when comparing keys
+      orig.replace(s"{{$patt}}", attributes(patt))
+    }
+
+    val newLogUrlMap = if (allPatterns.contains("FILE_NAME")) {
+      // allAttributeKeys should contain "LOG_FILES"
+      attributes("LOG_FILES").split(",").map { file =>
+        file -> updatedUrl.replace("{{FILE_NAME}}", file)
+      }.toMap
+    } else {
+      Map("log" -> updatedUrl)
+    }
+
+    replaceExecutorLogs(exec, newLogUrlMap)
+  }
+
+  private def logFailToRenewLogUrls(
+      reason: String,
+      allPatterns: Set[String],
+      allAttributes: Set[String]): Unit = {
+    if (informedForMissingAttributes.compareAndSet(false, true)) {
+      logInfo(s"Fail to renew executor log urls: $reason. Required: $allPatterns / " +
+        s"available: $allAttributes. Falling back to show app's original log urls.")
+    }
+  }
+
+  private def replaceExecutorLogs(
+      source: v1.ExecutorSummary,
+      newExecutorLogs: Map[String, String]): v1.ExecutorSummary = {
+    new v1.ExecutorSummary(source.id, source.hostPort, source.isActive, source.rddBlocks,
+      source.memoryUsed, source.diskUsed, source.totalCores, source.maxTasks, source.activeTasks,
+      source.failedTasks, source.completedTasks, source.totalTasks, source.totalDuration,
+      source.totalGCTime, source.totalInputBytes, source.totalShuffleRead,
+      source.totalShuffleWrite, source.isBlacklisted, source.maxMemory, source.addTime,
+      source.removeTime, source.removeReason, newExecutorLogs, source.memoryMetrics,
+      source.blacklistedInStages, source.peakMemoryMetrics, source.attributes)
+  }
+
+}
+
+private[spark] object HistoryAppStatusStore {
+  val CUSTOM_URL_PATTERN_REGEX: Regex = "\\{\\{([A-Za-z0-9_\\-]+)\\}\\}".r
+}

core/src/main/scala/org/apache/spark/executor/CoarseGrainedExecutorBackend.scala

@@ -60,7 +60,8 @@ private[spark] class CoarseGrainedExecutorBackend(
     rpcEnv.asyncSetupEndpointRefByURI(driverUrl).flatMap { ref =>
       // This is a very fast action so we can use "ThreadUtils.sameThread"
       driver = Some(ref)
-      ref.ask[Boolean](RegisterExecutor(executorId, self, hostname, cores, extractLogUrls))
+      ref.ask[Boolean](RegisterExecutor(executorId, self, hostname, cores, extractLogUrls,
+        extractAttributes))
     }(ThreadUtils.sameThread).onComplete {
       // This is a very fast action so we can use "ThreadUtils.sameThread"
       case Success(msg) =>
@@ -76,6 +77,12 @@ private[spark] class CoarseGrainedExecutorBackend(
       .map(e => (e._1.substring(prefix.length).toLowerCase(Locale.ROOT), e._2))
   }
 
+  def extractAttributes: Map[String, String] = {
+    val prefix = "SPARK_EXECUTOR_ATTRIBUTE_"
+    sys.env.filterKeys(_.startsWith(prefix))
+      .map(e => (e._1.substring(prefix.length).toUpperCase(Locale.ROOT), e._2))
+  }
+
   override def receive: PartialFunction[Any, Unit] = {
     case RegisteredExecutor =>
       logInfo("Successfully registered with driver")

core/src/main/scala/org/apache/spark/internal/config/History.scala

@@ -125,4 +125,22 @@ private[spark] object History {
   val KERBEROS_KEYTAB = ConfigBuilder("spark.history.kerberos.keytab")
     .stringConf
     .createOptional
+
+  val CUSTOM_EXECUTOR_LOG_URL = ConfigBuilder("spark.history.custom.executor.log.url")
+    .doc("Specifies custom spark executor log url for supporting external log service instead of " +
+      "using cluster managers' application log urls in the history server. Spark will support " +
+      "some path variables via patterns which can vary on cluster manager. Please check the " +
+      "documentation for your cluster manager to see which patterns are supported, if any. " +
+      "This configuration has no effect on a live application, it only affects the history server.")
+    .stringConf
+    .createOptional
+
+  val APPLY_CUSTOM_EXECUTOR_LOG_URL_TO_INCOMPLETE_APP =
+    ConfigBuilder("spark.history.custom.executor.log.url.applyIncompleteApplication")
+      .doc("Whether to apply custom executor log url, as specified by " +
+        "`spark.history.custom.executor.log.url`, to incomplete application as well. " +
+        "Even if this is true, this still only affects the behavior of the history server, " +
+        "not running spark applications.")
+      .booleanConf
+      .createWithDefault(true)
 }

core/src/main/scala/org/apache/spark/scheduler/SchedulerBackend.scala

@@ -69,6 +69,13 @@ private[spark] trait SchedulerBackend {
    */
   def getDriverLogUrls: Option[Map[String, String]] = None
 
+  /**
+   * Get the attributes on driver. These attributes are used to replace log URLs when
+   * custom log url pattern is specified.
+   * @return Map containing attributes on driver.
+   */
+  def getDriverAttributes: Option[Map[String, String]] = None
+
   /**
    * Get the max number of tasks that can be concurrent launched currently.
    * Note that please don't cache the value returned by this method, because the number can change

core/src/main/scala/org/apache/spark/scheduler/SparkListener.scala

@@ -192,7 +192,8 @@ case class SparkListenerApplicationStart(
     time: Long,
     sparkUser: String,
     appAttemptId: Option[String],
-    driverLogs: Option[Map[String, String]] = None) extends SparkListenerEvent
+    driverLogs: Option[Map[String, String]] = None,
+    driverAttributes: Option[Map[String, String]] = None) extends SparkListenerEvent
 
 @DeveloperApi
 case class SparkListenerApplicationEnd(time: Long) extends SparkListenerEvent

core/src/main/scala/org/apache/spark/scheduler/cluster/CoarseGrainedClusterMessage.scala

@@ -63,7 +63,8 @@ private[spark] object CoarseGrainedClusterMessages {
       executorRef: RpcEndpointRef,
       hostname: String,
       cores: Int,
-      logUrls: Map[String, String])
+      logUrls: Map[String, String],
+      attributes: Map[String, String])
     extends CoarseGrainedClusterMessage
 
   case class StatusUpdate(executorId: String, taskId: Long, state: TaskState,

core/src/main/scala/org/apache/spark/scheduler/cluster/CoarseGrainedSchedulerBackend.scala

@@ -183,7 +183,7 @@ class CoarseGrainedSchedulerBackend(scheduler: TaskSchedulerImpl, val rpcEnv: Rp
 
     override def receiveAndReply(context: RpcCallContext): PartialFunction[Any, Unit] = {
 
-      case RegisterExecutor(executorId, executorRef, hostname, cores, logUrls) =>
+      case RegisterExecutor(executorId, executorRef, hostname, cores, logUrls, attributes) =>
         if (executorDataMap.contains(executorId)) {
           executorRef.send(RegisterExecutorFailed("Duplicate executor ID: " + executorId))
           context.reply(true)
@@ -207,7 +207,7 @@ class CoarseGrainedSchedulerBackend(scheduler: TaskSchedulerImpl, val rpcEnv: Rp
           totalCoreCount.addAndGet(cores)
           totalRegisteredExecutors.addAndGet(1)
           val data = new ExecutorData(executorRef, executorAddress, hostname,
-            cores, cores, logUrls)
+            cores, cores, logUrls, attributes)
           // This must be synchronized because variables mutated
           // in this block are read when requesting executors
           CoarseGrainedSchedulerBackend.this.synchronized {

core/src/main/scala/org/apache/spark/scheduler/cluster/ExecutorData.scala

@@ -34,5 +34,6 @@ private[cluster] class ExecutorData(
    override val executorHost: String,
    var freeCores: Int,
    override val totalCores: Int,
-   override val logUrlMap: Map[String, String]
-) extends ExecutorInfo(executorHost, totalCores, logUrlMap)
+   override val logUrlMap: Map[String, String],
+   override val attributes: Map[String, String]
+) extends ExecutorInfo(executorHost, totalCores, logUrlMap, attributes)