Github user squito commented on a diff in the pull request:
https://github.com/apache/spark/pull/8760#discussion_r42263885
--- Diff:
core/src/test/scala/org/apache/spark/scheduler/BlacklistTrackerSuite.scala ---
@@ -0,0 +1,155 @@
+/*
+ * 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 org.apache.spark.SparkFunSuite
+import org.scalatest.BeforeAndAfter
+import org.apache.spark.SparkConf
+import org.apache.spark.TaskEndReason
+import org.apache.spark.Success
+import org.apache.spark.ExceptionFailure
+import org.apache.spark.LocalSparkContext
+import org.apache.spark.SparkContext
+import org.mockito.Mockito.{when,spy}
+import org.apache.spark.util.ManualClock
+
+class BlacklistTrackerSuite extends SparkFunSuite with BeforeAndAfter with
LocalSparkContext {
+
+ val FAILURE: TaskEndReason = new ExceptionFailure(
+ "Fake",
+ "fake failure",
+ Array.empty[StackTraceElement],
+ "fake stack trace",
+ None,
+ None)
+
+ // Variable name can indicate basic information of taskInfo
+ // The format is "taskInfo_executorId_taskId_hostName"
+ val taskInfo_1_1_hostA = new TaskInfo(1L,1,1,0L,"1","hostA",
TaskLocality.ANY, false)
+ val taskInfo_1_2_hostA = new TaskInfo(2L,1,1,0L,"1","hostA",
TaskLocality.ANY, false)
+ val taskInfo_2_1_hostA = new TaskInfo(1L,1,1,0L,"2","hostA",
TaskLocality.ANY, false)
+ val taskInfo_2_2_hostA = new TaskInfo(2L,1,1,0L,"2","hostA",
TaskLocality.ANY, false)
+ val taskInfo_3_3_hostB = new TaskInfo(3L,1,1,0L,"3","hostB",
TaskLocality.ANY, false)
+
+ val clock = new ManualClock(0)
+
+ test ("expireExecutorsInBlacklist works") {
+ // expire time is set to 6s
+ val conf = new SparkConf().setAppName("test").setMaster("local")
+ .set("spark.ui.enabled","false")
+ .set("spark.scheduler.executorTaskBlacklistTime", "6000")
+
+ sc = new SparkContext(conf)
+ val scheduler = new TaskSchedulerImpl(sc, 1)
+
+ val tracker = new BlacklistTracker(conf, clock)
+ // Executor 1 into blacklist at Time 00:00:00
+ tracker.updateFailureExecutors(taskInfo_1_1_hostA, FAILURE)
+ assert(tracker.executorBlacklist(scheduler, 1L) === Set("1"))
+
+ clock.setTime(2000)
+ tracker.expireExecutorsInBlackList()
+ assert(tracker.executorBlacklist(scheduler, 1L) === Set("1"))
+ // Executor 1 failed again at Time 00::00:02
+ tracker.updateFailureExecutors(taskInfo_1_1_hostA, FAILURE)
+
+ clock.setTime(3000)
+ // Executor 2 failed at Time 00:00:03
+ tracker.updateFailureExecutors(taskInfo_2_1_hostA, FAILURE)
+ assert(tracker.executorBlacklist(scheduler, 1L) === Set("1", "2"))
+
+ clock.setTime(6000)
+ tracker.expireExecutorsInBlackList()
+ assert(tracker.executorBlacklist(scheduler, 1L) === Set("1", "2"))
+
+ clock.setTime(8000)
+ tracker.expireExecutorsInBlackList()
+ assert(tracker.executorBlacklist(scheduler, 1L) === Set("2"))
+
+ clock.setTime(10000)
+ tracker.expireExecutorsInBlackList()
+ assert(tracker.executorBlacklist(scheduler, 1L) === Set())
+ }
+
+ test("default strategy works") {
+ val conf = new SparkConf().setAppName("test").setMaster("local")
+ .set("spark.ui.enabled","false")
+ sc = new SparkContext(conf)
+ val scheduler = new TaskSchedulerImpl(sc, 1)
+
+ val tracker = new BlacklistTracker(conf, clock)
+ tracker.updateFailureExecutors(taskInfo_1_1_hostA, FAILURE)
+ tracker.updateFailureExecutors(taskInfo_2_1_hostA, FAILURE)
+ assert(tracker.executorBlacklist(scheduler, 1L) === Set("1","2"))
+ assert(tracker.executorBlacklist(scheduler, 2L) === Set())
+
+ tracker.updateFailureExecutors(taskInfo_1_1_hostA, Success)
+ assert(tracker.executorBlacklist(scheduler, 1L) === Set("2"))
+ assert(tracker.nodeBlacklist() === Set())
+
+ tracker.updateFailureExecutors(taskInfo_3_3_hostB, Success)
+ assert(tracker.executorBlacklist(scheduler, 1L) === Set("2"))
+
+ tracker.updateFailureExecutors(taskInfo_3_3_hostB, FAILURE)
+ assert(tracker.executorBlacklist(scheduler, 3L) === Set("3"))
+ assert(tracker.nodeBlacklist() === Set())
+
+ tracker.updateFailureExecutors(taskInfo_2_1_hostA, Success)
+ assert(tracker.executorBlacklist(scheduler, 1L) === Set())
+ }
+
+ test ("threshold strategy works") {
+ val conf = new SparkConf().setAppName("test").setMaster("local")
+ .set("spark.ui.enabled","false")
+ .set("spark.scheduler.blacklist.strategy", "threshold")
+ .set("spark.scheduler.blacklist.threshold.maxFailureTaskNumber", "1")
+ .set("spark.scheduler.blacklist.threshold.maxBlackExecutorNumber",
"0")
+
+ sc = new SparkContext(conf)
+ val scheduler = spy(new TaskSchedulerImpl(sc, 1))
+
when(scheduler.getExecutorsAliveOnHost("hostA")).thenReturn(Some(Set("1","2")))
+
+ val tracker = new BlacklistTracker(conf, clock)
+ tracker.updateFailureExecutors(taskInfo_1_1_hostA, FAILURE)
+ assert(tracker.executorBlacklist(scheduler, 1L) === Set())
+ tracker.updateFailureExecutors(taskInfo_1_2_hostA, FAILURE)
+ assert(tracker.executorBlacklist(scheduler, 2L) === Set("1", "2"))
+ assert(tracker.nodeBlacklist() === Set("hostA"))
--- End diff --
Sorry this is a pain, but I'd like you to make a few changes to increase
test coverage a bit:
(a) add executor "3" to hostA
(b) change "maxBlackExecutorNumber" to 1
(c) after failing two tasks on executor 1, check that all tasks are
blacklisted on executor 1, (regardless of task id):
```scala
tracker.updateFailureExecutors(taskInfo_1_1_hostA, FAILURE)
assert(tracker.executorBlacklist(scheduler, 1L) === Set())
tracker.updateFailureExecutors(taskInfo_1_1_hostA, FAILURE)
// executor 1 is now blacklisted from *all* tasks
(0 until 10).foreach { assert(tracker.executorBlacklist(scheduler, _) ===
Set("1")) }
assert(tracker.nodeBlacklist() === Set())
```
(d) then fail add more task failures, so that you blacklist executor 2,
which will trigger the node blacklist so also blacklist executor 3.
(e) after removing executor 1 from the blacklist, then for all tasks, the
blacklist is just executor 2 (executor 3 is also removed).
the point is to check that the separation of executor & node blacklists.
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
