[jira] [Commented] (SPARK-28699) Cache an indeterminate RDD could lead to incorrect result while stage rerun
[
https://issues.apache.org/jira/browse/SPARK-28699?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16910994#comment-16910994
]
Dongjoon Hyun commented on SPARK-28699:
---
Thank you for the update, [~XuanYuan]!
> Cache an indeterminate RDD could lead to incorrect result while stage rerun
> ---
>
> Key: SPARK-28699
> URL: https://issues.apache.org/jira/browse/SPARK-28699
> Project: Spark
> Issue Type: Bug
> Components: Spark Core
>Affects Versions: 2.3.3, 3.0.0, 2.4.3
>Reporter: Yuanjian Li
>Priority: Blocker
> Labels: correctness
>
> It's another case for the indeterminate stage/RDD rerun while stage rerun
> happened.
> We can reproduce this by the following code, thanks to Tyson for reporting
> this!
>
> {code:scala}
> import scala.sys.process._
> import org.apache.spark.TaskContext
> val res = spark.range(0, 1 * 1, 1).map{ x => (x % 1000, x)}
> // kill an executor in the stage that performs repartition(239)
> val df = res.repartition(113).cache.repartition(239).map { x =>
> if (TaskContext.get.attemptNumber == 0 && TaskContext.get.partitionId < 1 &&
> TaskContext.get.stageAttemptNumber == 0) {
> throw new Exception("pkill -f -n java".!!)
> }
> x
> }
> val r2 = df.distinct.count()
> {code}
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-28699) Cache an indeterminate RDD could lead to incorrect result while stage rerun
[
https://issues.apache.org/jira/browse/SPARK-28699?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16910990#comment-16910990
]
Yuanjian Li commented on SPARK-28699:
-
[~dongjoon] Sure, the affects version is spark-2.1 after 2.1.4, spark 2.2 after
2.2.3, spark 2.3 and spark 2.4, Jira field update is done.
> Cache an indeterminate RDD could lead to incorrect result while stage rerun
> ---
>
> Key: SPARK-28699
> URL: https://issues.apache.org/jira/browse/SPARK-28699
> Project: Spark
> Issue Type: Bug
> Components: Spark Core
>Affects Versions: 3.0.0
>Reporter: Yuanjian Li
>Priority: Blocker
> Labels: correctness
>
> It's another case for the indeterminate stage/RDD rerun while stage rerun
> happened. In the CachedRDDBuilder.
> We can reproduce this by the following code, thanks to Tyson for reporting
> this!
>
> {code:scala}
> import scala.sys.process._
> import org.apache.spark.TaskContext
> val res = spark.range(0, 1 * 1, 1).map{ x => (x % 1000, x)}
> // kill an executor in the stage that performs repartition(239)
> val df = res.repartition(113).cache.repartition(239).map { x =>
> if (TaskContext.get.attemptNumber == 0 && TaskContext.get.partitionId < 1 &&
> TaskContext.get.stageAttemptNumber == 0) {
> throw new Exception("pkill -f -n java".!!)
> }
> x
> }
> val r2 = df.distinct.count()
> {code}
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-28699) Cache an indeterminate RDD could lead to incorrect result while stage rerun
[
https://issues.apache.org/jira/browse/SPARK-28699?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16910987#comment-16910987
]
Dongjoon Hyun commented on SPARK-28699:
---
:)
BTW, I updated this to `Blocker` according to [~smilegator]'s advice.
> Cache an indeterminate RDD could lead to incorrect result while stage rerun
> ---
>
> Key: SPARK-28699
> URL: https://issues.apache.org/jira/browse/SPARK-28699
> Project: Spark
> Issue Type: Bug
> Components: Spark Core
>Affects Versions: 3.0.0
>Reporter: Yuanjian Li
>Priority: Blocker
> Labels: correctness
>
> It's another case for the indeterminate stage/RDD rerun while stage rerun
> happened. In the CachedRDDBuilder.
> We can reproduce this by the following code, thanks to Tyson for reporting
> this!
>
> {code:scala}
> import scala.sys.process._
> import org.apache.spark.TaskContext
> val res = spark.range(0, 1 * 1, 1).map{ x => (x % 1000, x)}
> // kill an executor in the stage that performs repartition(239)
> val df = res.repartition(113).cache.repartition(239).map { x =>
> if (TaskContext.get.attemptNumber == 0 && TaskContext.get.partitionId < 1 &&
> TaskContext.get.stageAttemptNumber == 0) {
> throw new Exception("pkill -f -n java".!!)
> }
> x
> }
> val r2 = df.distinct.count()
> {code}
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-28699) Cache an indeterminate RDD could lead to incorrect result while stage rerun
[
https://issues.apache.org/jira/browse/SPARK-28699?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16910983#comment-16910983
]
Kazuaki Ishizaki commented on SPARK-28699:
--
[~dongjoon] Thank you for pointing out my typo. You are right. I should have
said {2.3.4-rc1}.
Actually, while I was doing the following, it is not reflected at the
repository yet! After fixing this, let me restart the release process for
{2.3.4-rc1}.
{code}
Release details:
BRANCH: branch-2.3
VERSION: 2.3.4
TAG: v2.3.4-rc1
{code}
> Cache an indeterminate RDD could lead to incorrect result while stage rerun
> ---
>
> Key: SPARK-28699
> URL: https://issues.apache.org/jira/browse/SPARK-28699
> Project: Spark
> Issue Type: Bug
> Components: Spark Core
>Affects Versions: 3.0.0
>Reporter: Yuanjian Li
>Priority: Major
> Labels: correctness
>
> Related with SPARK-23207 SPARK-23243
> It's another case for the indeterminate stage/RDD rerun while stage rerun
> happened. In the CachedRDDBuilder.
> We can reproduce this by the following code, thanks to Tyson for reporting
> this!
>
> {code:scala}
> import scala.sys.process._
> import org.apache.spark.TaskContext
> val res = spark.range(0, 1 * 1, 1).map\{ x => (x % 1000, x)}
> // kill an executor in the stage that performs repartition(239)
> val df = res.repartition(113).cache.repartition(239).map { x =>
> if (TaskContext.get.attemptNumber == 0 && TaskContext.get.partitionId < 1 &&
> TaskContext.get.stageAttemptNumber == 0) {
> throw new Exception("pkill -f -n java".!!)
> }
> x
> }
> val r2 = df.distinct.count()
> {code}
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-28699) Cache an indeterminate RDD could lead to incorrect result while stage rerun
[
https://issues.apache.org/jira/browse/SPARK-28699?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16910974#comment-16910974
]
Dongjoon Hyun commented on SPARK-28699:
---
? [~kiszk]. `2.4.4-rc1` is `branch-2.4` and mine. You should not remove that.
I guess you wanted to say `2.3.4-rc1` and `2.3.4-rc1` is not created yet.
> Cache an indeterminate RDD could lead to incorrect result while stage rerun
> ---
>
> Key: SPARK-28699
> URL: https://issues.apache.org/jira/browse/SPARK-28699
> Project: Spark
> Issue Type: Bug
> Components: Spark Core
>Affects Versions: 3.0.0
>Reporter: Yuanjian Li
>Priority: Major
> Labels: correctness
>
> Related with SPARK-23207 SPARK-23243
> It's another case for the indeterminate stage/RDD rerun while stage rerun
> happened. In the CachedRDDBuilder, we miss tracking the `isOrderSensitive`
> characteristic to the newly created MapPartitionsRDD.
> We can reproduce this by the following code, thanks to Tyson for reporting
> this!
>
> {code:scala}
> import scala.sys.process._
> import org.apache.spark.TaskContext
> val res = spark.range(0, 1 * 1, 1).map\{ x => (x % 1000, x)}
> // kill an executor in the stage that performs repartition(239)
> val df = res.repartition(113).cache.repartition(239).map { x =>
> if (TaskContext.get.attemptNumber == 0 && TaskContext.get.partitionId < 1 &&
> TaskContext.get.stageAttemptNumber == 0) {
> throw new Exception("pkill -f -n java".!!)
> }
> x
> }
> val r2 = df.distinct.count()
> {code}
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-28699) Cache an indeterminate RDD could lead to incorrect result while stage rerun
[
https://issues.apache.org/jira/browse/SPARK-28699?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16910907#comment-16910907
]
Kazuaki Ishizaki commented on SPARK-28699:
--
[~smilegator] Thank you for cc. I wait for fixing this.
I was in the middle of releasing RC1. Thus, there is already {{2.4.4-rc1}} tag
in the [branch-2.3|https://github.com/apache/spark/tree/branch-2.3]. Should I
remove this tag and release rc1? Or should I leave this tag and release rc2 at
first?
> Cache an indeterminate RDD could lead to incorrect result while stage rerun
> ---
>
> Key: SPARK-28699
> URL: https://issues.apache.org/jira/browse/SPARK-28699
> Project: Spark
> Issue Type: Bug
> Components: Spark Core
>Affects Versions: 3.0.0
>Reporter: Yuanjian Li
>Priority: Major
> Labels: correctness
>
> Related with SPARK-23207 SPARK-23243
> It's another case for the indeterminate stage/RDD rerun while stage rerun
> happened. In the CachedRDDBuilder, we miss tracking the `isOrderSensitive`
> characteristic to the newly created MapPartitionsRDD.
> We can reproduce this by the following code, thanks to Tyson for reporting
> this!
>
> {code:scala}
> import scala.sys.process._
> import org.apache.spark.TaskContext
> val res = spark.range(0, 1 * 1, 1).map\{ x => (x % 1000, x)}
> // kill an executor in the stage that performs repartition(239)
> val df = res.repartition(113).cache.repartition(239).map { x =>
> if (TaskContext.get.attemptNumber == 0 && TaskContext.get.partitionId < 1 &&
> TaskContext.get.stageAttemptNumber == 0) {
> throw new Exception("pkill -f -n java".!!)
> }
> x
> }
> val r2 = df.distinct.count()
> {code}
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-28699) Cache an indeterminate RDD could lead to incorrect result while stage rerun
[
https://issues.apache.org/jira/browse/SPARK-28699?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16910859#comment-16910859
]
Xiao Li commented on SPARK-28699:
-
Also cc [~kiszk] Let us wait for this before starting RC1 for 2.3
> Cache an indeterminate RDD could lead to incorrect result while stage rerun
> ---
>
> Key: SPARK-28699
> URL: https://issues.apache.org/jira/browse/SPARK-28699
> Project: Spark
> Issue Type: Bug
> Components: Spark Core
>Affects Versions: 3.0.0
>Reporter: Yuanjian Li
>Priority: Major
> Labels: correctness
>
> Related with SPARK-23207 SPARK-23243
> It's another case for the indeterminate stage/RDD rerun while stage rerun
> happened. In the CachedRDDBuilder, we miss tracking the `isOrderSensitive`
> characteristic to the newly created MapPartitionsRDD.
> We can reproduce this by the following code, thanks to Tyson for reporting
> this!
>
> {code:scala}
> import scala.sys.process._
> import org.apache.spark.TaskContext
> val res = spark.range(0, 1 * 1, 1).map\{ x => (x % 1000, x)}
> // kill an executor in the stage that performs repartition(239)
> val df = res.repartition(113).cache.repartition(239).map { x =>
> if (TaskContext.get.attemptNumber == 0 && TaskContext.get.partitionId < 1 &&
> TaskContext.get.stageAttemptNumber == 0) {
> throw new Exception("pkill -f -n java".!!)
> }
> x
> }
> val r2 = df.distinct.count()
> {code}
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-28699) Cache an indeterminate RDD could lead to incorrect result while stage rerun
[
https://issues.apache.org/jira/browse/SPARK-28699?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16910795#comment-16910795
]
Dongjoon Hyun commented on SPARK-28699:
---
Hi, [~XuanYuan].
Could you check old Spark versions and update `Affects Version/s:` of this JIRA
issue?
> Cache an indeterminate RDD could lead to incorrect result while stage rerun
> ---
>
> Key: SPARK-28699
> URL: https://issues.apache.org/jira/browse/SPARK-28699
> Project: Spark
> Issue Type: Bug
> Components: Spark Core
>Affects Versions: 3.0.0
>Reporter: Yuanjian Li
>Priority: Major
> Labels: correctness
>
> Related with SPARK-23207 SPARK-23243
> It's another case for the indeterminate stage/RDD rerun while stage rerun
> happened. In the CachedRDDBuilder, we miss tracking the `isOrderSensitive`
> characteristic to the newly created MapPartitionsRDD.
> We can reproduce this by the following code, thanks to Tyson for reporting
> this!
>
> {code:scala}
> import scala.sys.process._
> import org.apache.spark.TaskContext
> val res = spark.range(0, 1 * 1, 1).map\{ x => (x % 1000, x)}
> // kill an executor in the stage that performs repartition(239)
> val df = res.repartition(113).cache.repartition(239).map { x =>
> if (TaskContext.get.attemptNumber == 0 && TaskContext.get.partitionId < 1 &&
> TaskContext.get.stageAttemptNumber == 0) {
> throw new Exception("pkill -f -n java".!!)
> }
> x
> }
> val r2 = df.distinct.count()
> {code}
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
[jira] [Commented] (SPARK-28699) Cache an indeterminate RDD could lead to incorrect result while stage rerun
[
https://issues.apache.org/jira/browse/SPARK-28699?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16905250#comment-16905250
]
Yuanjian Li commented on SPARK-28699:
-
The current [approach|https://github.com/apache/spark/pull/25420] just a
bandage fix for returning the wrong answer.
After we finish the work of indeterminate stage rerunning(SPARK-25341), we can
fix this by unpersisting the original RDD and rerunning the cached
indeterminate stage. Gives a preview codebase
[here|https://github.com/xuanyuanking/spark/tree/SPARK-28699-RERUN].
> Cache an indeterminate RDD could lead to incorrect result while stage rerun
> ---
>
> Key: SPARK-28699
> URL: https://issues.apache.org/jira/browse/SPARK-28699
> Project: Spark
> Issue Type: Bug
> Components: Spark Core
>Affects Versions: 3.0.0
>Reporter: Yuanjian Li
>Priority: Major
>
> Related with SPARK-23207 SPARK-23243
> It's another case for the indeterminate stage/RDD rerun while stage rerun
> happened. In the CachedRDDBuilder, we miss tracking the `isOrderSensitive`
> characteristic to the newly created MapPartitionsRDD.
> We can reproduce this by the following code, thanks to Tyson for reporting
> this!
>
> {code:scala}
> import scala.sys.process._
> import org.apache.spark.TaskContext
> val res = spark.range(0, 1 * 1, 1).map\{ x => (x % 1000, x)}
> // kill an executor in the stage that performs repartition(239)
> val df = res.repartition(113).cache.repartition(239).map { x =>
> if (TaskContext.get.attemptNumber == 0 && TaskContext.get.partitionId < 1 &&
> TaskContext.get.stageAttemptNumber == 0) {
> throw new Exception("pkill -f -n java".!!)
> }
> x
> }
> val r2 = df.distinct.count()
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.14#76016)
-
To unsubscribe, e-mail: issues-unsubscr...@spark.apache.org
For additional commands, e-mail: issues-h...@spark.apache.org
