Github user squito commented on a diff in the pull request:
https://github.com/apache/spark/pull/15249#discussion_r81895140
--- Diff:
core/src/main/scala/org/apache/spark/scheduler/TaskSetBlacklist.scala ---
@@ -0,0 +1,136 @@
+/*
+ * 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.scheduler
+
+import scala.collection.mutable.{HashMap, HashSet}
+
+import org.apache.spark.SparkConf
+import org.apache.spark.internal.config
+import org.apache.spark.internal.Logging
+import org.apache.spark.util.Clock
+
+/**
+ * Handles blacklisting executors and nodes within a taskset. This
includes blacklisting specific
+ * (task, executor) / (task, nodes) pairs, and also completely
blacklisting executors and nodes
+ * for the entire taskset.
+ *
+ * THREADING: As a helper to [[TaskSetManager]], this class is designed
to only be called from code
+ * with a lock on the TaskScheduler (e.g. its event handlers). It should
not be called from other
+ * threads.
+ */
+private[scheduler] class TaskSetBlacklist(val conf: SparkConf, val
stageId: Int, val clock: Clock)
+ extends Logging {
+
+ private val MAX_TASK_ATTEMPTS_PER_EXECUTOR =
conf.get(config.MAX_TASK_ATTEMPTS_PER_EXECUTOR)
+ private val MAX_TASK_ATTEMPTS_PER_NODE =
conf.get(config.MAX_TASK_ATTEMPTS_PER_NODE)
+ private val MAX_FAILURES_PER_EXEC_STAGE =
conf.get(config.MAX_FAILURES_PER_EXEC_STAGE)
+ private val MAX_FAILED_EXEC_PER_NODE_STAGE =
conf.get(config.MAX_FAILED_EXEC_PER_NODE_STAGE)
+ private val TIMEOUT_MILLIS = BlacklistTracker.getBlacklistTimeout(conf)
+
+ /**
+ * A map from each executor to the task failures on that executor.
+ */
+ val execToFailures: HashMap[String, ExecutorFailuresInTaskSet] = new
HashMap()
+
+ /**
+ * Map from node to all executors on it with failures. Needed because
we want to know about
+ * executors on a node even after they have died.
+ */
+ private val nodeToExecsWithFailures: HashMap[String, HashSet[String]] =
new HashMap()
+ private val nodeToBlacklistedTasks: HashMap[String, HashSet[Int]] = new
HashMap()
+ private val blacklistedExecs: HashSet[String] = new HashSet()
+ private val blacklistedNodes: HashSet[String] = new HashSet()
+
+ /**
+ * Return true if this executor is blacklisted for the given task. This
does *not*
+ * need to return true if the executor is blacklisted for the entire
stage.
+ * That is to keep this method as fast as possible in the inner-loop of
the
+ * scheduler, where those filters will have already been applied.
+ */
+ def isExecutorBlacklistedForTask(
+ executorId: String,
+ index: Int): Boolean = {
+ execToFailures.get(executorId)
+ .map { execFailures =>
+ val count =
execFailures.taskToFailureCountAndExpiryTime.get(index).map(_._1).getOrElse(0)
+ count >= MAX_TASK_ATTEMPTS_PER_EXECUTOR
+ }
+ .getOrElse(false)
+ }
+
+ def isNodeBlacklistedForTask(
+ node: String,
+ index: Int): Boolean = {
+ nodeToBlacklistedTasks.get(node)
+ .map(_.contains(index))
+ .getOrElse(false)
+ }
+
+ /**
+ * Return true if this executor is blacklisted for the given stage.
Completely ignores whether
+ * anything to do with the node the executor is on. That
+ * is to keep this method as fast as possible in the inner-loop of the
scheduler, where those
+ * filters will already have been applied.
+ */
+ def isExecutorBlacklistedForTaskSet(executorId: String): Boolean = {
+ blacklistedExecs.contains(executorId)
+ }
+
+ def isNodeBlacklistedForTaskSet(node: String): Boolean = {
+ blacklistedNodes.contains(node)
+ }
--- End diff --
I know its verbose but I'd prefer to keep it. Especially once
application-level blacklisting is added
(https://github.com/apache/spark/pull/14079), there are lots of different
blacklisting checks and its easy to get confused. The class isn't always the
most prominent (esp in scala) so I think it helps a bit here.
---
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]
