executor: fix CTE may be blocked when query report error

[guo-shaoge]

Signed-off-by: guo-shaoge [email protected]

What problem does this PR solve?

Issue Number: close #31302

Problem Summary:

mysql> explain with recursive cte(iter, first, second, result) as 
                       (select 1, first, second, first+second from src union all
                        select iter+1, second, result, second+result from cte where iter < 80 )
                        select * from cte;
+-------------------------------------+
| id                                  |
+-------------------------------------+
| CTEFullScan_18                      |
| CTE_0                               |
| ├─Projection_11(Seed Part)          |
| │ └─TableReader_13                  |
| │   └─TableFullScan_12              |
| └─Projection_14(Recursive Part)     |
|   └─Selection_15                    |
|     └─CTETable_16                   |
+-------------------------------------+

Why blocked:

1. Projection_14 will return overflow error
2. ProjectionExec.Close() will be called, and it will wait for projectionInputFetcher to finish.
3. projectionInputFetcher is blocked because CTETable_16 is blocked in CTETable_16.Next() because CTETable_16 is waiting for e.iterInTbl.GetBegCh().
4. CTE_0 will reset iterInTbl.begCh when error occurs, so begCh is reset, but nobody close begCh anymore. So CTETable_16 is blocked on <- e.iterInTbl.GetBegCh()

Why needs begCh:

1. Some executors(like IndexLookUpJoin) start workers in Open() so it will call child.Next() in Open() phase.
2. But data in iterInTbl is not ready in Open() phase, so these executor will read empty chunk which is unexpected.
3. That's why we need begCh, it indicates when data is ready. When computeSeedPart() or each iteration in CTEExec is done, iterInTbl will be readable, we will close(begCh) and CTETableExec will continue to return correct chunk.

But it's too complicated and caused this block bug. So I decide to change code in IndexLookUpJoin and remove begCh.

What is changed and how it works?

For IndexLookUpJoin, IndexNestedLookUpHashJoin and IndexLookUpMergeJoin, start works in Next() instead of Open().

Check List

Tests

• Unit test
• Integration test
• Manual test (add detailed scripts or steps below)
• No code

Side effects

• Performance regression: There maybe performance loss because workers in executor start in the first call of Next().
• Performance regression: Consumes more Memory
• Breaking backward compatibility

Documentation

• Affects user behaviors
• Contains syntax changes
• Contains variable changes
• Contains experimental features
• Changes MySQL compatibility

Release note

executor: fix CTE is block when query report error

[ti-chi-bot]

[REVIEW NOTIFICATION]

This pull request has been approved by:

• XuHuaiyu
• wjhuang2016

To complete the pull request process, please ask the reviewers in the list to review by filling /cc @reviewer in the comment.
After your PR has acquired the required number of LGTMs, you can assign this pull request to the committer in the list by filling /assign @committer in the comment to help you merge this pull request.

The full list of commands accepted by this bot can be found here.

Reviewer can indicate their review by submitting an approval review.
Reviewer can cancel approval by submitting a request changes review.

[sre-bot]

Code Coverage Details: https://codecov.io/github/pingcap/tidb/commit/8acb1e0d1660464ae536e74f45a70fa56b531d28

[guo-shaoge]

/run-mysql-test

[guo-shaoge]

/cc @wjhuang2016

[guo-shaoge]

/run-mysql-test

[guo-shaoge]

/run-mysql-test

[guo-shaoge]

/run-mysql-test

[XuHuaiyu]

LGTM

[guo-shaoge]

/merge

[ti-chi-bot]

This pull request has been accepted and is ready to merge.

Commit hash: a83cfe0

[ti-srebot]

cherry pick to release-5.1 in PR #33185

[ti-srebot]

cherry pick to release-5.2 in PR #33187

[ti-srebot]

cherry pick to release-5.3 in PR #33189

[ti-srebot]

cherry pick to release-5.4 in PR #33190