ddl: schedule as many jobs as possible in one round, simplify sql to query jobs #54438

D3Hunter · 2024-07-04T06:53:39Z

What problem does this PR solve?

Issue Number: ref #54436

Problem Summary:

What changed and how does it work?

schedule as many jobs as possible in one round, simplify sql to query jobs
replace 2 ddl test hook with failpoint
rename methods

Check List

Tests

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

before this pr: create 100K tables under same database on owner on 1pd/3tidb/3tikv using 10 threads, run 2 time, takes 31m16/29m5. avg cpu about [335, 400]

after: run 3 times, takes 32m8/22m56/25m29, avg cpu about [200, 280]. the test result is not stable as expected, seems there are other components interfering DDL process, will check it later. but overall it it's better, both in time and resource usage

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

Release note

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

None

tiprow · 2024-07-04T06:54:04Z

Hi @D3Hunter. 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.

pkg/ddl/job_table.go

codecov · 2024-07-04T07:13:24Z

Codecov Report

Attention: Patch coverage is 80.41237% with 19 lines in your changes missing coverage. Please review.

Project coverage is 56.3967%. Comparing base (be16d49) to head (1edc320).
Report is 29 commits behind head on master.

Additional details and impacted files

@@                Coverage Diff                @@
##             master     #54438         +/-   ##
=================================================
- Coverage   74.7851%   56.3967%   -18.3884%     
=================================================
  Files          1536       1666        +130     
  Lines        361822     614408     +252586     
=================================================
+ Hits         270589     346506      +75917     
- Misses        71592     244564     +172972     
- Partials      19641      23338       +3697

Flag	Coverage Δ
integration	`37.2223% <63.9175%> (?)`
unit	`72.1366% <77.3195%> (-1.5610%)`	⬇️

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

Components	Coverage Δ
dumpling	`52.9656% <ø> (-2.2339%)`	⬇️
parser	`∅ <ø> (∅)`
br	`53.0896% <ø> (+5.1741%)`	⬆️

D3Hunter · 2024-07-05T03:11:30Z

@lance6716

tangenta · 2024-07-05T03:14:30Z

pkg/ddl/ddl_worker_test.go

@@ -75,7 +73,7 @@ func TestAddBatchJobError(t *testing.T) {
 }

 func TestParallelDDL(t *testing.T) {
-	store, dom := testkit.CreateMockStoreAndDomainWithSchemaLease(t, testLease)
+	store, _ := testkit.CreateMockStoreAndDomainWithSchemaLease(t, testLease)


Suggested change

store, _ := testkit.CreateMockStoreAndDomainWithSchemaLease(t, testLease)

store := testkit.CreateMockStoreWithSchemaLease(t, testLease)

pkg/ddl/job_table.go

lance6716 · 2024-07-05T07:10:04Z

pkg/ddl/job_table.go

+		if wk == nil {
+			// should not happen, we have checked available() before, and we are
+			// the only routine consumes worker.
+			logutil.DDLLogger().Info("no %v worker available now", zap.Stringer("type", targetPool.tp()))


no need to use %v

lance6716 · 2024-07-05T07:11:45Z

pkg/ddl/job_table.go

-	failpoint.InjectCall("beforeRunJobWithWorker")
+// runOneJobStep runs one step of the DDL job. we are using online-schema-change,
+// one job might go through multiple steps, each step is one job state change such
+// as from 'done' -> 'synced, or one schema state change such as 'delete only' -> 'write only'.


Suggested change

// as from 'done' -> 'synced, or one schema state change such as 'delete only' -> 'write only'.

// as from 'done' -> 'synced', or one schema state change such as 'delete only' -> 'write only'.

lance6716 · 2024-07-05T07:15:11Z

pkg/ddl/job_table.go

@@ -641,14 +633,6 @@ func (s *jobScheduler) runJobWithWorker(wk *worker, job *model.Job) error {
 	return nil
 }

-func (*jobScheduler) markJobProcessing(se *sess.Session, job *model.Job) error {


Maybe involve another DDL dev to think about that we don't persist it.

ti-chi-bot · 2024-07-05T07:17:01Z

[LGTM Timeline notifier]

Timeline:

2024-07-05 05:49:08.862621375 +0000 UTC m=+1562675.348110207: ☑️ agreed by tangenta.
2024-07-05 07:17:00.773897714 +0000 UTC m=+1567947.259386546: ☑️ agreed by lance6716.

ti-chi-bot · 2024-07-05T08:11:30Z

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: hawkingrei, lance6716, tangenta

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~~ [hawkingrei,lance6716,tangenta]
~~pkg/ddl/OWNERS~~ [tangenta]
~~pkg/statistics/OWNERS~~ [hawkingrei]

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

D3Hunter · 2024-07-05T08:12:28Z

/hold

D3Hunter · 2024-07-09T03:34:14Z

/unhold

D3Hunter · 2024-07-09T04:10:52Z

/retest

tiprow · 2024-07-09T04:11:20Z

@D3Hunter: 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.

D3Hunter added 2 commits July 4, 2024 14:18

change

d1ffea1

change

487cae7

ti-chi-bot bot added release-note-none Denotes a PR that doesn't merit a release note. size/XL Denotes a PR that changes 500-999 lines, ignoring generated files. labels Jul 4, 2024

lance6716 reviewed Jul 4, 2024

View reviewed changes

pkg/ddl/job_table.go Outdated Show resolved Hide resolved

D3Hunter changed the title ~~ddl: schedule as many jobs as possible in one round~~ ddl: schedule as many jobs as possible in one round, simplify sql to query jobs Jul 4, 2024

change

9d1e7d8

ti-chi-bot bot added the sig/planner SIG: Planner label Jul 4, 2024

D3Hunter mentioned this pull request Jul 4, 2024

ddl code refactor/optimize #54436

Open

55 tasks

tangenta reviewed Jul 5, 2024

View reviewed changes

fix comments

07aaffb

tangenta reviewed Jul 5, 2024

View reviewed changes

pkg/ddl/job_table.go Show resolved Hide resolved

tangenta approved these changes Jul 5, 2024

View reviewed changes

ti-chi-bot bot added the needs-1-more-lgtm Indicates a PR needs 1 more LGTM. label Jul 5, 2024

lance6716 approved these changes Jul 5, 2024

View reviewed changes

ti-chi-bot bot added lgtm and removed needs-1-more-lgtm Indicates a PR needs 1 more LGTM. labels Jul 5, 2024

fix comments

1edc320

hawkingrei approved these changes Jul 5, 2024

View reviewed changes

ti-chi-bot bot added the approved label Jul 5, 2024

ti-chi-bot bot added the do-not-merge/hold Indicates that a PR should not merge because someone has issued a /hold command. label Jul 5, 2024

D3Hunter mentioned this pull request Jul 8, 2024

lightning: improve log #54483

Merged

13 tasks

ti-chi-bot bot removed the do-not-merge/hold Indicates that a PR should not merge because someone has issued a /hold command. label Jul 9, 2024

ti-chi-bot bot merged commit 8c1ce99 into pingcap:master Jul 9, 2024
21 checks passed

D3Hunter deleted the schedule-amap2 branch July 9, 2024 04:48

lance6716 mentioned this pull request Jul 10, 2024

ddl: fix check running ingest job #54543

Merged

13 tasks

D3Hunter mentioned this pull request Jul 11, 2024

infoschema: use maps.Clone to replace iterating #54581

Merged

13 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

ddl: schedule as many jobs as possible in one round, simplify sql to query jobs #54438

ddl: schedule as many jobs as possible in one round, simplify sql to query jobs #54438

D3Hunter commented Jul 4, 2024 •

edited

Loading

tiprow bot commented Jul 4, 2024

codecov bot commented Jul 4, 2024 •

edited

Loading

D3Hunter commented Jul 5, 2024

tangenta Jul 5, 2024

lance6716 Jul 5, 2024

lance6716 Jul 5, 2024

lance6716 Jul 5, 2024

ti-chi-bot bot commented Jul 5, 2024

ti-chi-bot bot commented Jul 5, 2024

D3Hunter commented Jul 5, 2024

D3Hunter commented Jul 9, 2024

D3Hunter commented Jul 9, 2024

tiprow bot commented Jul 9, 2024

	store, _ := testkit.CreateMockStoreAndDomainWithSchemaLease(t, testLease)
	store := testkit.CreateMockStoreWithSchemaLease(t, testLease)

	// as from 'done' -> 'synced, or one schema state change such as 'delete only' -> 'write only'.
	// as from 'done' -> 'synced', or one schema state change such as 'delete only' -> 'write only'.

ddl: schedule as many jobs as possible in one round, simplify sql to query jobs #54438

ddl: schedule as many jobs as possible in one round, simplify sql to query jobs #54438

Conversation

D3Hunter commented Jul 4, 2024 • edited Loading

What problem does this PR solve?

What changed and how does it work?

Check List

Release note

tiprow bot commented Jul 4, 2024

codecov bot commented Jul 4, 2024 • edited Loading

Codecov Report

D3Hunter commented Jul 5, 2024

tangenta Jul 5, 2024

Choose a reason for hiding this comment

lance6716 Jul 5, 2024

Choose a reason for hiding this comment

lance6716 Jul 5, 2024

Choose a reason for hiding this comment

lance6716 Jul 5, 2024

Choose a reason for hiding this comment

ti-chi-bot bot commented Jul 5, 2024

[LGTM Timeline notifier]

ti-chi-bot bot commented Jul 5, 2024

D3Hunter commented Jul 5, 2024

D3Hunter commented Jul 9, 2024

D3Hunter commented Jul 9, 2024

tiprow bot commented Jul 9, 2024

D3Hunter commented Jul 4, 2024 •

edited

Loading

codecov bot commented Jul 4, 2024 •

edited

Loading