[SPARK-29042][Core][BRANCH-2.4] Sampling-based RDD with unordered input should be INDETERMINATE

[viirya]

What changes were proposed in this pull request?

We already have found and fixed the correctness issue before when RDD output is INDETERMINATE. One missing part is sampling-based RDD. This kind of RDDs is order sensitive to its input. A sampling-based RDD with unordered input, should be INDETERMINATE.

Note that this is backport of original PR to branch-2.4.

Why are the changes needed?

A sampling-based RDD with unordered input is just like MapPartitionsRDD with isOrderSensitive parameter as true. The RDD output can be different after a rerun.

It is a problem in ML applications.

In ML, sample is used to prepare training data. ML algorithm fits the model based on the sampled data. If rerun tasks of sample produce different output during model fitting, ML results will be unreliable and also buggy.

Each sample is random output, but once you sampled, the output should be determinate.

Does this PR introduce any user-facing change?

Previously, a sampling-based RDD can possibly come with different output after a rerun.
After this patch, sampling-based RDD is INDETERMINATE. For an INDETERMINATE map stage, currently Spark scheduler will re-try all the tasks of the failed stage.

How was this patch tested?

Added test.

[viirya]

cc @cloud-fan @gatorsmile

[HyukjinKwon]

LGTM if tests pass

[SparkQA]

Test build #110859 has finished for PR 25826 at commit e17dd66.

• This patch fails due to an unknown error code, -9.
• This patch merges cleanly.
• This patch adds no public classes.

[cloud-fan]

retest this please

[SparkQA]

Test build #110885 has finished for PR 25826 at commit e17dd66.

• This patch fails Spark unit tests.
• This patch merges cleanly.
• This patch adds no public classes.

[srowen]

Looks OK pending tests.

[HyukjinKwon]

retest this please

[SparkQA]

Test build #110915 has finished for PR 25826 at commit e17dd66.

• This patch passes all tests.
• This patch merges cleanly.
• This patch adds no public classes.

[viirya]

Thanks! Merged to 2.4.

[dongjoon-hyun]

Ur, @viirya .
Did you ignore the merge script recommendation? The merge script will give you a recommendation after cleaning up the hidden text. If the recommendation looks still ugly, you can stop the merging and polish the PR description first by removing the hidden text. Then, the script will not bother you about that~

commit efcca57d3db4e706ada40ed4bd16f2b1512c3532 (HEAD -> branch-2.4, apache/branch-2.4)
Author: Liang-Chi Hsieh <[email protected]>
Date:   Wed Sep 18 13:05:29 2019 -0700

    [SPARK-29042][CORE][BRANCH-2.4] Sampling-based RDD with unordered input should be INDETERMINATE

    <!--
    Thanks for sending a pull request!  Here are some tips for you:
      1. If this is your first time, please read our contributor guidelines: https://spark.apache.org/contributing.html
      2. Ensure you have added or run the appropriate tests for your PR: https://spark.apache.org/developer-tools.html
      3. If the PR is unfinished, add '[WIP]' in your PR title, e.g., '[WIP][SPARK-XXXX] Your PR title ...'.
      4. Be sure to keep the PR description updated to reflect all changes.
      5. Please write your PR title to summarize what this PR proposes.
      6. If possible, provide a concise example to reproduce the issue for a faster review.
    -->

    ### What changes were proposed in this pull request?
    <!--
    Please clarify what changes you are proposing. The purpose of this section is to outline the changes and how this PR fixes the issue.
    If possible, please consider writing useful notes for better and faster reviews in your PR. See the examples below.
      1. If you refactor some codes with changing classes, showing the class hierarchy will help reviewers.
      2. If you fix some SQL features, you can provide some references of other DBMSes.
      3. If there is design documentation, please add the link.
      4. If there is a discussion in the mailing list, please add the link.
    -->

    We already have found and fixed the correctness issue before when RDD output is INDETERMINATE. One missing part is sampling-based RDD. This kind of RDDs is order sensitive to its input. A sampling-based RDD with unordered input, should be INDETERMINATE.

    Note that this is backport of original PR to branch-2.4.

    ### Why are the changes needed?
    <!--
    Please clarify why the changes are needed. For instance,
      1. If you propose a new API, clarify the use case for a new API.
      2. If you fix a bug, you can clarify why it is a bug.
    -->

[viirya]

Did you ignore the merge script recommendation? The merge script will give you a recommendation after cleaning up the hidden text. If the recommendation looks still ugly, you can stop the merging and policy the PR description first by removing the hidden text. Then, the script will not bother you about that~

I noticed this. I recall there is recommendation for PR title, but don't remember I saw the recommendation for description.

Oh, I see. At merging this, my repo is based on branch-2.4. And the script dev/merge_spark_pr.py in branch-2.4 does not help modify the PR description...

[dongjoon-hyun]

Oh.. I got it. I had better backport them to branch-2.4, too. Thanks!