[SPARK-2947] DAGScheduler resubmit the stage into an infinite loop by witgo · Pull Request #1877 · apache/spark

witgo · 2014-08-10T15:05:47Z

No description provided.

SparkQA · 2014-08-10T15:09:39Z

QA tests have started for PR 1877. This patch merges cleanly.
View progress: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/18281/consoleFull

SparkQA · 2014-08-10T16:00:17Z

QA results for PR 1877:
- This patch PASSES unit tests.
- This patch merges cleanly
- This patch adds no public classes

For more information see test ouptut:
https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/18281/consoleFull

mateiz · 2014-08-10T23:00:40Z

@witgo can you explain how this happens and why the fix works, and add a unit test for it? We can't really merge something like this without a test.

witgo · 2014-08-11T10:19:59Z

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

@mateiz
There is no cancel running tasks in the stage . When any one of the running tasks which throws an exception.The following code will be repeated.

failedStages += failedStage failedStages += mapStage

Stage will be unnecessary resubmit by resubmitFailedStages

witgo · 2014-08-11T16:44:52Z

It takes some time to add a test for this.

SparkQA · 2014-08-12T08:34:44Z

QA tests have started for PR 1877. This patch merges cleanly.
View progress: https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/18367/consoleFull

SparkQA · 2014-08-12T09:26:29Z

QA results for PR 1877:
- This patch PASSES unit tests.
- This patch merges cleanly
- This patch adds no public classes

For more information see test ouptut:
https://amplab.cs.berkeley.edu/jenkins/job/SparkPullRequestBuilder/18367/consoleFull

davies · 2014-08-25T20:38:56Z

@witgo Could you rebase this PR onto master? There are some conflict right now.

SparkQA · 2014-08-26T03:01:18Z

QA tests have started for PR 1877 at commit 3484c29.

This patch merges cleanly.

SparkQA · 2014-08-26T03:58:36Z

QA tests have finished for PR 1877 at commit 3484c29.

This patch passes unit tests.
This patch merges cleanly.
This patch adds the following public classes (experimental):
- $FWDIR/bin/spark-submit --class org.apache.spark.repl.Main "$
- $FWDIR/bin/spark-submit --class org.apache.spark.repl.Main "$
- class KMeansModel (val clusterCenters: Array[Vector]) extends Serializable
- class BoundedFloat(float):
- class JoinedRow2 extends Row
- class JoinedRow3 extends Row
- class JoinedRow4 extends Row
- class JoinedRow5 extends Row
- class GenericRow(protected[sql] val values: Array[Any]) extends Row
- abstract class MutableValue extends Serializable
- final class MutableInt extends MutableValue
- final class MutableFloat extends MutableValue
- final class MutableBoolean extends MutableValue
- final class MutableDouble extends MutableValue
- final class MutableShort extends MutableValue
- final class MutableLong extends MutableValue
- final class MutableByte extends MutableValue
- final class MutableAny extends MutableValue
- final class SpecificMutableRow(val values: Array[MutableValue]) extends MutableRow
- case class CountDistinct(expressions: Seq[Expression]) extends PartialAggregate
- case class CollectHashSet(expressions: Seq[Expression]) extends AggregateExpression
- case class CollectHashSetFunction(
- case class CombineSetsAndCount(inputSet: Expression) extends AggregateExpression
- case class CombineSetsAndCountFunction(
- case class CountDistinctFunction(
- case class MaxOf(left: Expression, right: Expression) extends Expression
- case class NewSet(elementType: DataType) extends LeafExpression
- case class AddItemToSet(item: Expression, set: Expression) extends Expression
- case class CombineSets(left: Expression, right: Expression) extends BinaryExpression
- case class CountSet(child: Expression) extends UnaryExpression
- case class ExplainCommand(plan: LogicalPlan, extended: Boolean = false) extends Command

witgo · 2014-08-28T05:54:11Z

@rxin could you take a look at this PR? Thanks!

rxin · 2014-08-28T05:55:39Z

Can you explain what problem you are seeing?

SparkQA · 2014-08-28T05:56:05Z

QA tests have started for PR 1877 at commit c4b0f91.

This patch merges cleanly.

witgo · 2014-08-28T06:08:00Z

SPARK-3224 is the same problem.
This PR adds some boundary judgments and removed some redundant code

kayousterhout · 2014-08-28T06:46:49Z

core/src/test/scala/org/apache/spark/scheduler/DAGSchedulerSuite.scala

can you change this and/or the name for the test at line 438? They are currently almost identical such that it's unclear what the point of each test is.

SparkQA · 2014-08-28T06:51:29Z

QA tests have finished for PR 1877 at commit c4b0f91.

This patch passes unit tests.
This patch merges cleanly.
This patch adds no public classes.

SparkQA · 2014-08-29T15:59:10Z

QA tests have started for PR 1877 at commit bf6f81a.

This patch merges cleanly.

SparkQA · 2014-08-29T16:59:11Z

QA tests have finished for PR 1877 at commit bf6f81a.

This patch passes unit tests.
This patch merges cleanly.
This patch adds no public classes.

witgo · 2014-09-09T14:21:07Z

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

@rxin Because there is no cancel running tasks in the stage. stage.pendingTasks.contains(task) is necessary.

witgo · 2014-09-09T14:26:18Z

screenshots:

SparkQA · 2014-09-13T05:01:16Z

QA tests have started for PR 1877 at commit bf6f81a.

This patch merges cleanly.

rxin · 2014-09-13T05:11:24Z

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

once you put this within the conditional statement, only one executor failure will be handled for each stage. that means if there are two executor fails, the 2nd one gets ignored by the dagscheduler, isn't it?

@rxin Yes, here is unnecessary modifications to processing logic, I negligence.
是的,这里处理逻辑被不必要的修改了,疏忽了.

SparkQA · 2014-09-13T05:52:15Z

QA tests have finished for PR 1877 at commit bf6f81a.

This patch fails unit tests.
This patch merges cleanly.
This patch adds the following public classes (experimental):
- throw new IllegalStateException("The main method in the given main class must be static")
- class TaskCompletionListenerException(errorMessages: Seq[String]) extends Exception
- class Dummy(object):

SparkQA · 2014-09-14T02:59:26Z

QA tests have started for PR 1877 at commit 958d7db.

This patch merges cleanly.

SparkQA · 2014-09-14T04:04:47Z

QA tests have finished for PR 1877 at commit 958d7db.

This patch passes unit tests.
This patch merges cleanly.
This patch adds the following public classes (experimental):
- class RatingDeserializer(FramedSerializer):
- class Encoder[T <: NativeType](columnType: NativeColumnType[T]) extends compression.Encoder[T]
- class Encoder[T <: NativeType](columnType: NativeColumnType[T]) extends compression.Encoder[T]
- class Encoder[T <: NativeType](columnType: NativeColumnType[T]) extends compression.Encoder[T]
- class Encoder extends compression.Encoder[IntegerType.type]
- class Decoder(buffer: ByteBuffer, columnType: NativeColumnType[IntegerType.type])
- class Encoder extends compression.Encoder[LongType.type]
- class Decoder(buffer: ByteBuffer, columnType: NativeColumnType[LongType.type])

…der (apache#1877) Parquet footer metadata is now always read twice in vectorized parquet reader. When the NameNode is under high pressure, it will cost time to read twice. Actually we can avoid reading the footer twice by reading all row groups in advance and filter row groups according to filters that require push down (no need to read the footer metadata again the second time). Reduce the reading of footer in vectorized parquet reader no existing tests Closes apache#39950 from yabola/skip_footer. Authored-by: chenliang.lu <marssss2929@gmail.com> Signed-off-by: Chao Sun <sunchao@apple.com> Co-authored-by: chenliang.lu <marssss2929@gmail.com>

witgo reviewed Aug 11, 2014
View reviewed changes

witgo changed the title ~~[WIP][SPARK-2947] DAGScheduler resubmit the stage into an infinite loop~~ [SPARK-2947] DAGScheduler resubmit the stage into an infinite loop Aug 12, 2014

witgo force-pushed the SPARK-2947 branch from e3172d7 to 3484c29 Compare August 26, 2014 02:56

witgo force-pushed the SPARK-2947 branch from 3484c29 to c4b0f91 Compare August 28, 2014 05:52

kayousterhout reviewed Aug 28, 2014
View reviewed changes

witgo force-pushed the SPARK-2947 branch 2 times, most recently from 089577f to bf6f81a Compare August 29, 2014 15:55

witgo reviewed Sep 9, 2014
View reviewed changes

rxin reviewed Sep 13, 2014
View reviewed changes

DAGScheduler resubmit the stage into an infinite loop

3250f9f

review commit

958d7db

witgo force-pushed the SPARK-2947 branch from bf6f81a to 958d7db Compare September 14, 2014 02:55

witgo closed this Dec 3, 2014

Conversation

witgo commented Aug 10, 2014

Uh oh!

SparkQA commented Aug 10, 2014

Uh oh!

SparkQA commented Aug 10, 2014

Uh oh!

mateiz commented Aug 10, 2014

Uh oh!

witgo Aug 11, 2014

Choose a reason for hiding this comment

Uh oh!

witgo commented Aug 11, 2014

Uh oh!

SparkQA commented Aug 12, 2014

Uh oh!

SparkQA commented Aug 12, 2014

Uh oh!

davies commented Aug 25, 2014

Uh oh!

SparkQA commented Aug 26, 2014

Uh oh!

SparkQA commented Aug 26, 2014

Uh oh!

witgo commented Aug 28, 2014

Uh oh!

rxin commented Aug 28, 2014

Uh oh!

SparkQA commented Aug 28, 2014

Uh oh!

witgo commented Aug 28, 2014

Uh oh!

kayousterhout Aug 28, 2014

Choose a reason for hiding this comment

Uh oh!

SparkQA commented Aug 28, 2014

Uh oh!

SparkQA commented Aug 29, 2014

Uh oh!

SparkQA commented Aug 29, 2014

Uh oh!

witgo Sep 9, 2014

Choose a reason for hiding this comment

Uh oh!

witgo commented Sep 9, 2014

Uh oh!

SparkQA commented Sep 13, 2014

Uh oh!

rxin Sep 13, 2014

Choose a reason for hiding this comment

Uh oh!

witgo Sep 13, 2014

Choose a reason for hiding this comment

Uh oh!

SparkQA commented Sep 13, 2014

Uh oh!

SparkQA commented Sep 14, 2014

Uh oh!

SparkQA commented Sep 14, 2014

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants