global sort: reduce the memory usage of merge sort concurrent reader

[fzzf678]

What problem does this PR solve?

Issue Number: close #62853

Problem Summary:

What changed and how does it work?

Reduce the concurrency of concurrent readers, pass the outerConcurrency correctly.

In the previous PR, we increased the concurrency to get better performance and if there are multiple readers concurrently read, the outerConcurrency will be used to control the whole concurrency. https://github.com/pingcap/tidb/pull/48871/files#diff-1ea5d87a202f7a3a089796fa8ac98c520879a2bdfdc739ed98a2261a9a7e8ad1R377-R381

But the outerConcurrency was forgotten to be passed(see https://github.com/pingcap/tidb/pull/48871/files#diff-b573df9af3de0ee789f4ec5114d897cc2f9b5ac57f7cc19757e2353127680edfR91), the reader concurrency is always 256 no matter how many reader we use, this leads to the OOM.

Check List

Tests

• Unit test
• Integration test
• Manual test (add detailed scripts or steps below)

• For the OOM problem, after we pass the outerConcurrency correctly in this PR, the add index can be success without OOM. The profile as shown below, we can see the memory occupied by switchToConcurrentReader is 2GiB, this is expected because there are 8 readers and each has 32 concurrency with 256MiB memory, totally 8*256MiB = 2GiB.

• About the merge sort performance, we test the merge sort time using tidb_ddl_reorg_worker_cnt = 4. Before this PR, each subtask tasks about 44-46 minutes. After this PR, each subtask tasks about 33-34 minutes to finish merge sort.
  We detect the reason is, previously the concurrency was too high, which slowed down the pull of some goroutines, causing tail requests to drag down the overall situation. After reducing the concurrency, resource contention and congestion were alleviated, slow requests were reduced, and the overall tail delay converged. Resource contention is reduced or congestion handling capabilities are improved, making data transmission more stable and efficient. This PR is mainly to fix the OOM problem, maybe we can continue investigating performance improvements later.

Before(repeat 3 times):

+---------------------+---------------------+----------+------+-----------+---------+
| start_time          | end_time            | duration | step | exec_id   | state   |
+---------------------+---------------------+----------+------+-----------+---------+
| 2025-08-12 05:28:30 | 2025-08-12 06:14:54 | 00:46:24 |    2 | tc-tidb-0 | succeed |
| 2025-08-12 05:28:30 | 2025-08-12 06:14:25 | 00:45:55 |    2 | tc-tidb-2 | succeed |
| 2025-08-12 05:28:30 | 2025-08-12 06:13:26 | 00:44:56 |    2 | tc-tidb-1 | succeed |
+---------------------+---------------------+----------+------+-----------+---------+

+---------------------+---------------------+----------+------+-----------+---------+
| start_time          | end_time            | duration | step | exec_id   | state   |
+---------------------+---------------------+----------+------+-----------+---------+
| 2025-08-12 08:41:37 | 2025-08-12 09:27:12 | 00:45:35 |    2 | tc-tidb-0 | succeed |
| 2025-08-12 08:41:37 | 2025-08-12 09:27:03 | 00:45:26 |    2 | tc-tidb-1 | succeed |
| 2025-08-12 08:41:37 | 2025-08-12 09:26:24 | 00:44:47 |    2 | tc-tidb-2 | succeed |
+---------------------+---------------------+----------+------+-----------+---------+

+---------------------+---------------------+----------+------+-----------+---------+
| start_time          | end_time            | duration | step | exec_id   | state   |
+---------------------+---------------------+----------+------+-----------+---------+
| 2025-08-12 10:48:09 | 2025-08-12 11:33:27 | 00:45:18 |    2 | tc-tidb-1 | succeed |
| 2025-08-12 10:48:09 | 2025-08-12 11:33:24 | 00:45:15 |    2 | tc-tidb-2 | succeed |
| 2025-08-12 10:48:10 | 2025-08-12 11:33:21 | 00:45:11 |    2 | tc-tidb-0 | succeed |
+---------------------+---------------------+----------+------+-----------+---------+

After(repeat 3 times):

+---------------------+---------------------+----------+------+-------------------------------------------------------------------------+---------+
| start_time          | end_time            | duration | step | exec_id                                                                 | state   |
+---------------------+---------------------+----------+------+-------------------------------------------------------------------------+---------+
| 2025-08-19 01:40:47 | 2025-08-19 02:15:01 | 00:34:14 |    2 | tc-tidb-1 | succeed |
| 2025-08-19 01:40:47 | 2025-08-19 02:14:25 | 00:33:38 |    2 | tc-tidb-0 | succeed |
| 2025-08-19 01:40:47 | 2025-08-19 02:14:10 | 00:33:23 |    2 | tc-tidb-2 | succeed |
+---------------------+---------------------+----------+------+-------------------------------------------------------------------------+---------+

+---------------------+---------------------+----------+------+-------------------------------------------------------------------------+---------+
| start_time          | end_time            | duration | step | exec_id                                                                 | state   |
+---------------------+---------------------+----------+------+-------------------------------------------------------------------------+---------+
| 2025-08-19 03:47:08 | 2025-08-19 04:21:05 | 00:33:57 |    2 | tc-tidb-0 | succeed |
| 2025-08-19 03:47:08 | 2025-08-19 04:20:58 | 00:33:50 |    2 | tc-tidb-1 | succeed |
| 2025-08-19 03:47:08 | 2025-08-19 04:20:42 | 00:33:34 |    2 | tc-tidb-2 | succeed |
+---------------------+---------------------+----------+------+-------------------------------------------------------------------------+---------+

+---------------------+---------------------+----------+------+-------------------------------------------------------------------------+---------+
| start_time          | end_time            | duration | step | exec_id                                                                 | state   |
+---------------------+---------------------+----------+------+-------------------------------------------------------------------------+---------+
| 2025-08-19 12:42:42 | 2025-08-19 13:17:04 | 00:34:22 |    2 | tc-tidb-2 | succeed |
| 2025-08-19 12:42:42 | 2025-08-19 13:16:44 | 00:34:02 |    2 | tc-tidb-0 | succeed |
| 2025-08-19 12:42:42 | 2025-08-19 13:16:41 | 00:33:59 |    2 | tc-tidb-1 | succeed |
+---------------------+---------------------+----------+------+-------------------------------------------------------------------------+---------+

• No need to test

  • I checked and no code files have been changed.

Side effects

• Performance regression: Consumes more CPU
• 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

Please refer to Release Notes Language Style Guide to write a quality release note.

None

[ti-chi-bot]

Skipping CI for Draft Pull Request.
If you want CI signal for your change, please convert it to an actual PR.
You can still manually trigger a test run with /test all

[tiprow]

Hi @fzzf678. Thanks for your PR.

PRs from untrusted users cannot be marked as trusted with /ok-to-test in this repo meaning untrusted PR authors can never trigger tests themselves. Collaborators can still trigger tests on the PR using /test all.

I understand the commands that are listed here.

Details

Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes-sigs/prow repository.

[codecov]

Codecov Report

❌ Patch coverage is 66.66667% with 1 line in your changes missing coverage. Please review.
✅ Project coverage is 73.4498%. Comparing base (6361402) to head (aee5212).
⚠️ Report is 472 commits behind head on master.

Additional details and impacted files

@@               Coverage Diff                @@
##             master     #62921        +/-   ##
================================================
+ Coverage   72.8138%   73.4498%   +0.6359%     
================================================
  Files          1823       1853        +30     
  Lines        494575     502708      +8133     
================================================
+ Hits         360119     369238      +9119     
+ Misses       112573     110903      -1670     
- Partials      21883      22567       +684     

Flag	Coverage Δ
integration	44.7680% <0.0000%> (?)
unit	72.2793% <66.6666%> (+0.0052%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Components	Coverage Δ
dumpling	52.8700% <ø> (ø)
parser	∅ <ø> (∅)
br	46.3505% <ø> (+0.0178%)	⬆️

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[fzzf678]

tidb/pkg/ddl/backfilling_merge_sort.go

Line 92 in 16ec21b

memSizePerCon := res.Mem.Capacity() / res.CPU.Capacity()

Should we consider using this memSizePerConto control the concurrency of concurrent reader?
The memSizePerCon can't control, need another value to control the total memory usage of concurrent reader

[D3Hunter]

before this PR, each subtask tasks about 44-46 minutes. After this PR, each subtask tasks about 33-34 minutes to finish merge sort.

should before this PR OOM? where the 44-46 minutes came from?

i think this PR only avoid OOM, why the performance is getting better?

[fzzf678]

before this PR, each subtask tasks about 44-46 minutes. After this PR, each subtask tasks about 33-34 minutes to finish merge sort.

should before this PR OOM? where the 44-46 minutes came from?

i think this PR only avoid OOM, why the performance is getting better?

The OOM occurs when tidb_ddl_reorg_worker_cnt = 8, the 44-46 minutes is tested by using the tidb_ddl_reorg_worker_cnt = 4. Supply this to the description.

[ti-chi-bot]

[LGTM Timeline notifier]

Timeline:

• 2025-08-20 03:44:24.556505674 +0000 UTC m=+411472.499681190: ☑️ agreed by wjhuang2016.
• 2025-08-21 02:50:50.857473549 +0000 UTC m=+494658.800649066: ☑️ agreed by D3Hunter.

[hawkingrei]

/retest

[ti-chi-bot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: D3Hunter, tangenta, wjhuang2016

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

The pull request process is described here

Details

Needs approval from an approver in each of these files:

• pkg/lightning/OWNERS [D3Hunter]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[ti-chi-bot]

In response to a cherrypick label: new pull request created to branch release-8.5: #64744.
But this PR has conflicts, please resolve them!