store/copr: optimize copIterator by avoid start new goroutine

[crazycs520]

What problem does this PR solve?

Issue Number: ref #56649

Problem Summary: optimize cop iterator by avoid start new goroutine

What changed and how does it work?

Before this PR, copIterator always start at least 1 copIteratorWorker goroutine and 1 copIteratorTaskSender goroutine. That's too wasteful for small queries.

This PR avoid start new goroutine for small queries, for better performance, following is the sysbench oltp_index_scan test, and the range size is 2:

sysbench --config-file=sysbench.conf oltp_index_scan --tables=16 --table-size=1000000 --threads=100 --range-size=2 run

The example query is like this:

SELECT
  k
FROM
  sbtest1
WHERE
  k BETWEEN ?
  AND ? [arguments: (503925, 503926)]

Which Plan is:

| id                 | estRows | estCost | actRows | task      | access object               | execution info                                                                                                                                                                                                                                                                                         | operator info                           | memory    | disk  |
| IndexReader_6      | 2.07    | 31.20   | 2       | root      |                             | time:1.31ms, loops:2, cop_task: {num: 1, max: 1.15ms, proc_keys: 2, tot_proc: 81.2µs, tot_wait: 575.8µs, copr_cache: disabled, build_task_duration: 12.2µs, max_distsql_concurrency: 1}, rpc_info:{Cop:{num_rpc:1, total_time:1.12ms}}                                                                 | index:IndexRangeScan_5                  | 278 Bytes | N/A   |
| └─IndexRangeScan_5 | 2.07    | 336.86  | 2       | cop[tikv] | table:sbtest1, index:k_1(k) | tikv_task:{time:0s, loops:1}, scan_detail: {total_process_keys: 2, total_process_keys_size: 92, total_keys: 3, get_snapshot_time: 539.6µs, rocksdb: {key_skipped_count: 2, block: {cache_hit_count: 6}}}, time_detail: {total_process_time: 81.2µs, total_wait_time: 575.8µs, tikv_wall_time: 751.3µs} | range:[503925,503926], keep order:false | N/A       | N/A   |

version	workload	thread	QPS	QPS Increase
master	oltp_index_scan	10	9224
This PR	oltp_index_scan	10	9471	2.6%
master	oltp_index_scan	50	36072
This PR	oltp_index_scan	50	37392	3.6%
master	oltp_index_scan	100	41977
This PR	oltp_index_scan	100	43554	3.7%
master	oltp_index_scan	150	42785
This PR	oltp_index_scan	150	46430	8.5%

Check List

Tests

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

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

[tiprow]

Hi @crazycs520. 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.

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

Attention: Patch coverage is 88.18182% with 26 lines in your changes missing coverage. Please review.

Project coverage is 73.6562%. Comparing base (aec0fc5) to head (fdeff5b).
Report is 26 commits behind head on master.

Additional details and impacted files

@@               Coverage Diff                @@
##             master     #57522        +/-   ##
================================================
+ Coverage   73.2097%   73.6562%   +0.4465%     
================================================
  Files          1679       1679                
  Lines        462339     462466       +127     
================================================
+ Hits         338477     340635      +2158     
+ Misses       103082     101054      -2028     
+ Partials      20780      20777         -3     

Flag	Coverage Δ
integration	43.3960% <75.9090%> (?)
unit	72.3521% <88.1818%> (+0.0077%)	⬆️

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

Components	Coverage Δ
dumpling	52.6910% <ø> (ø)
parser	∅ <ø> (∅)
br	46.0380% <ø> (+0.0060%)	⬆️

[crazycs520]

/retest-required

[tiprow]

@crazycs520: Cannot trigger testing until a trusted user reviews the PR and leaves an /ok-to-test message.

In response to this:

/retest-required

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.

[crazycs520]

/retest-required

[tiprow]

@crazycs520: Cannot trigger testing until a trusted user reviews the PR and leaves an /ok-to-test message.

In response to this:

/retest-required

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.

[you06]

I'm inclined that the TryCopLiteWorker is a request level option.

Like an index lookup with 1 row result set. If TryCopLiteWorker is a session level option, it's only enabled once in index scan, and in table lookup, the atomic.CompareAndSwapUint32(tryCopLiteWorker, 0, 1) will fail and execute with multi-goroutine model.

[crazycs520]

I'm afraid that make TryCopLiteWorker in request level option may have similar problem in this comment, there is a known case and I have added test for it. And I'm not sure if there are other cases like this, so I prefer keep TryCopLiteWorker to be statement level.

As index lookup with 1 row result set, you are right, but because of the above problems, I don't have a good way to fix this.

[crazycs520]

/retest-required

[you06]

LGTM, once we have async kv interface, we can remove TryCopLiteWorker and make more coprocessor tasks benefit from this opimization.
Wait for @cfzjywxk's approve.

[ti-chi-bot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: cfzjywxk, zyguan

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

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [cfzjywxk,zyguan]
• pkg/distsql/OWNERS [cfzjywxk]

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]

[LGTM Timeline notifier]

Timeline:

• 2024-12-16 04:16:23.878897798 +0000 UTC m=+843973.967700342: ☑️ agreed by zyguan.
• 2024-12-17 02:05:29.661253228 +0000 UTC m=+922519.750055771: ☑️ agreed by cfzjywxk.

[cfzjywxk]

/retest

[tiprow]

@cfzjywxk: Cannot trigger testing until a trusted user reviews the PR and leaves an /ok-to-test message.

In response to this:

/retest

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.

[cfzjywxk]

/ok-to-test

[cfzjywxk]

/retest

[crazycs520]

/retest-required