[runners-spark] Do not set accTimestamp to null in SparkCombineFn (#28256)

[je-ik]

Fixes #28256

---

Thank you for your contribution! Follow this checklist to help us incorporate your contribution quickly and easily:

• Mention the appropriate issue in your description (for example: addresses #123), if applicable. This will automatically add a link to the pull request in the issue. If you would like the issue to automatically close on merging the pull request, comment fixes #<ISSUE NUMBER> instead.
• Update CHANGES.md with noteworthy changes.
• If this contribution is large, please file an Apache Individual Contributor License Agreement.

See the Contributor Guide for more tips on how to make review process smoother.

To check the build health, please visit https://github.com/apache/beam/blob/master/.test-infra/BUILD_STATUS.md

GitHub Actions Tests Status (on master branch)

See CI.md for more information about GitHub Actions CI or the workflows README to see a list of phrases to trigger workflows.

[github-actions]

Checks are failing. Will not request review until checks are succeeding. If you'd like to override that behavior, comment assign set of reviewers

[aromanenko-dev]

Run Spark ValidatesRunner

[aromanenko-dev]

Run Spark Runner Tpcds Tests

[aromanenko-dev]

Run Spark Runner Nexmark Tests

[github-actions]

Assigning reviewers. If you would like to opt out of this review, comment assign to next reviewer:

R: @lostluck added as fallback since no labels match configuration

Available commands:

• stop reviewer notifications - opt out of the automated review tooling
• remind me after tests pass - tag the comment author after tests pass
• waiting on author - shift the attention set back to the author (any comment or push by the author will return the attention set to the reviewers)

The PR bot will only process comments in the main thread (not review comments).

[aromanenko-dev]

CC: @mosche @echauchot

[je-ik]

I'm not convinced this fix is 100% correct. I'd need an instance of a failing job to see what exactly causes the timestamp to be null.

[je-ik]

@aromanenko-dev I think I have the complete understanding. The condition in the assignment was wrong (don't have any clue how it was reasoned), the problem is that it was hidden, because until there was at least one call to add to the accumulator (which is nearly always the case), the timestamp was overwritten.

The bug manifested itself under the following conditions:
a) Combine.withHotKeyFanout, because that merges accumulators without any elements being added (in the postHotFanoutCombine phase)
b) TimestampCombiner.EARLIEST, because that can create accumulator timestamps of BoundedWindow.TIMESTAMP_MIN_VALUE.

Both conditions are satisfied in the case of query 3 in TPC-DS.

The combination is somewhat niche, which is why it was not caught by VR tests. We can add a VR test for this, but I think that the bug was so specific, that it makes little sense.

[je-ik]

Run Spark Runner Nexmark Tests

[je-ik]

Run Spark Runner Tpcds Tests

[je-ik]

Run Spark ValidatesRunner

[aromanenko-dev]

Run Spark Runner Tpcds Tests

[aromanenko-dev]

Great, LGTM, thanks a lot! It sounds reasonable for me

The condition in the assignment was wrong (don't have any clue how it was reasoned)

=) It was added 4+ years ago...

The combination is somewhat niche, which is why it was not caught by VR tests. We can add a VR test for this, but I think that the bug was so specific, that it makes little sense.

Actually, according to TPC-DS Grafana dashboard, it caused a lot of queries failed without catching it with VR tests, that worried me. TBH, I didn't see if any of users were affected (mostly because I guess SparkRunner is not so popular now) but it would be great to add any kind of regression test of this if possible.

Let's just wait for TPC-DS job results before merging.

[aromanenko-dev]

Run Spark Runner Tpcds Tests

[je-ik]

Added the test, it is surely better to have a test proving the fix, though in this case it is only "the code does not contain this bug". Nevertheless the line of thinking that introduced in could in theory reappear. 👍

[je-ik]

Run Spark Runner Tpcds Tests

[je-ik]

Run Java PreCommit

[je-ik]

Run Java PreCommit

[je-ik]

Run Spark ValidatesRunner