restore: use token bucket to balance download requests. #49887

3pointer · 2023-12-28T10:04:35Z

What problem does this PR solve?

Issue Number: ref #49886
Problem Summary:

introduce use token bucket(downloadV2) to balance download requests when granlurity is coarse-grained.

What changed and how does it work?

Check List

Tests

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

ti-chi-bot · 2023-12-28T10:04:38Z

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

sre-bot · 2023-12-28T10:04:43Z

All committers have signed the CLA.

tiprow · 2023-12-28T10:04:48Z

Hi @3pointer. 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/test-infra repository.

3pointer · 2023-12-28T10:05:18Z

Wait for https://github.com/pingcap/tidb/pull/48244/files merge first

3pointer · 2024-01-04T08:09:39Z

/ok-to-tests

codecov · 2024-01-04T08:24:27Z

Codecov Report

Merging #49887 (498160f) into master (9b0fd9e) will decrease coverage by 12.0160%.
Report is 8 commits behind head on master.
The diff coverage is 13.0281%.

Additional details and impacted files

@@                Coverage Diff                @@
##             master     #49887         +/-   ##
=================================================
- Coverage   79.3229%   67.3070%   -12.0160%     
=================================================
  Files          2447       2558        +111     
  Lines        673700     840148     +166448     
=================================================
+ Hits         534399     565479      +31080     
- Misses       117932     250900     +132968     
- Partials      21369      23769       +2400

Flag	Coverage Δ
integration	`36.6604% <12.6760%> (?)`
unit	`79.2826% <4.9295%> (-0.0404%)`	⬇️

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

Components	Coverage Δ
dumpling	`73.6130% <ø> (ø)`
parser	`∅ <ø> (∅)`
br	`71.7267% <13.0281%> (+3.2506%)`	⬆️

3pointer · 2024-01-04T11:03:40Z

/test pull-br-integration-test

ti-chi-bot · 2024-01-04T11:03:42Z

@3pointer: The specified target(s) for /test were not found.
The following commands are available to trigger required jobs:

/test build
/test check-dev
/test check-dev2
/test mysql-test
/test pull-integration-ddl-test
/test pull-lightning-integration-test
/test pull-mysql-client-test
/test unit-test

The following commands are available to trigger optional jobs:

/test canary-notify-when-compatibility-sections-changed
/test pingcap/tidb/canary_ghpr_unit_test
/test pull-br-integration-test
/test pull-common-test
/test pull-e2e-test
/test pull-integration-common-test
/test pull-integration-copr-test
/test pull-integration-jdbc-test
/test pull-integration-mysql-test
/test pull-integration-nodejs-test
/test pull-sqllogic-test
/test pull-tiflash-test

Use /test all to run the following jobs that were automatically triggered:

pingcap/tidb/ghpr_build
pingcap/tidb/ghpr_check
pingcap/tidb/ghpr_check2
pingcap/tidb/ghpr_mysql_test
pingcap/tidb/ghpr_unit_test
pingcap/tidb/pull_integration_ddl_test
pingcap/tidb/pull_mysql_client_test

In response to this:

/test pull-br-integration-test

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/test-infra repository.

tiprow · 2024-01-04T11:04:01Z

@3pointer: The specified target(s) for /test were not found.
The following commands are available to trigger required jobs:

/test tiprow_fast_test

Use /test all to run all jobs.

In response to this:

/test pull-br-integration-test

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/test-infra repository.

3pointer · 2024-01-04T12:48:25Z

/test pull-br-integration-test

tiprow · 2024-01-04T12:48:47Z

@3pointer: The specified target(s) for /test were not found.
The following commands are available to trigger required jobs:

/test tiprow_fast_test

Use /test all to run all jobs.

In response to this:

/test pull-br-integration-test

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/test-infra repository.

YuJuncen

rest lgtm

YuJuncen · 2024-01-05T07:36:23Z

br/pkg/restore/import.go

+	for _, sstMeta := range resultMetasMap {
+		downloadMetas = append(downloadMetas, sstMeta)
+	}
+	return downloadMetas, nil


Suggested change

for _, sstMeta := range resultMetasMap {

downloadMetas = append(downloadMetas, sstMeta)

}

return downloadMetas, nil

return maps.Values(resultMetasMap), nil

YuJuncen · 2024-01-05T07:37:42Z

br/pkg/restore/import.go

+				}
+
+				mu.Lock()
+				sstMeta, ok := downloadMetasMap[file.Name]


I will perfer make a small closure and use defer for unlocking.

I feel like is a tiny logic flaw, no need to make closure.

YuJuncen · 2024-01-05T07:38:32Z

br/pkg/restore/import.go

+				sstMeta, ok := downloadMetasMap[file.Name]
+				if !ok {
+					mu.Unlock()
+					return errors.New("not found file key for download sstMeta")


Please also provide the file name (perhaps also the content of the map) here. (Is this code really reachable? Maybe panic here is also acceptable.)

is unreachable. just a protect code.

YuJuncen · 2024-01-05T07:39:28Z

br/pkg/restore/import.go

+				}
+				var err error
+				var resp *import_sstpb.DownloadResponse
+				err = utils.WithRetry(ctx, func() error {


Maybe try utils.WithRetryV2, which allows us get the return value directly.

YuJuncen · 2024-01-05T07:41:53Z

br/pkg/restore/import.go

+					return errors.Annotate(berrors.ErrKVDownloadFailed, resp.GetError().GetMessage())
+				}
+				if resp.GetIsEmpty() {
+					return errors.Trace(berrors.ErrKVRangeIsEmpty)


Maybe also provide the range of the file for debugging.

we had log outside

log.Warn("download file skipped", logutil.Files(files), logutil.Region(info.Region), logutil.Key("startKey", startKey), logutil.Key("endKey", endKey), logutil.Key("file-simple-start", files[0].StartKey), logutil.Key("file-simple-end", files[0].EndKey),

It seems the file's key range won't be surely printed? We only print the first file there. Even in most case there should be only one file, querying log is usually harder than the direct error message.

ErrKVRangeIsEmpty is always a retryable error and never break the restoration. it appears somehow doesn't really means error. I think is kind of feature rather than error.

br/pkg/restore/import.go

br/pkg/restore/pipeline_items.go

br/tests/br_full/run.sh

Co-authored-by: 山岚 <[email protected]>

…into patch_per_store_concurrency

YuJuncen

rest lgtm

YuJuncen · 2024-01-05T08:57:49Z

br/pkg/restore/import.go

+					return errors.Annotate(berrors.ErrKVDownloadFailed, resp.GetError().GetMessage())
+				}
+				if resp.GetIsEmpty() {
+					return errors.Trace(berrors.ErrKVRangeIsEmpty)


It seems the file's key range won't be surely printed? We only print the first file there. Even in most case there should be only one file, querying log is usually harder than the direct error message.

YuJuncen · 2024-01-05T08:59:45Z

br/pkg/restore/import.go

+				sstMeta, ok := downloadMetasMap[file.Name]
+				if !ok {
+					mu.Unlock()
+					return errors.Errorf("not found file key for download sstMeta", file.Name)


It seems there isn't a printf placeholder in the format string.

3pointer · 2024-01-08T03:29:22Z

/retest

3pointer · 2024-01-08T04:33:28Z

/retest

Leavrth · 2024-01-08T04:00:48Z

br/pkg/restore/client.go

 	stores, err := conn.GetAllTiKVStoresWithRetry(ctx, rc.pdClient, util.SkipTiFlash)
 	if err != nil {
 		log.Fatal("failed to get stores", zap.Error(err))
 	}
-	concurrencyPerStore := 512
+	concurrencyPerStore := rc.GetConcurrencyPerStore()
 	for _, store := range stores {
-		ch := make(chan struct{}, concurrencyPerStore)
-		for i := 0; i < concurrencyPerStore; i += 1 {
-			ch <- struct{}{}
-		}
+		ch := utils.BuildWorkerTokenChannel(concurrencyPerStore)


actually only when useToKenBucket == true, the storeWorkerPoolMap need to be initialized.

yes, but I think it's okay to keep it here.

Leavrth · 2024-01-08T04:06:42Z

br/pkg/restore/client.go

 	rc.workerPool = utils.NewWorkerPool(c, "file")
+	rc.concurrencyPerStore = c


does it need to be removed?

br/pkg/restore/import.go

ti-chi-bot · 2024-01-08T07:36:02Z

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: Leavrth, YuJuncen

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:

~~br/OWNERS~~ [Leavrth,YuJuncen]

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 · 2024-01-08T07:36:05Z

[LGTM Timeline notifier]

Timeline:

2024-01-05 08:59:56.437246112 +0000 UTC m=+1786.021499800: ☑️ agreed by YuJuncen.
2024-01-08 07:36:04.745049656 +0000 UTC m=+255954.329303344: ☑️ agreed by Leavrth.

tiprow · 2024-01-08T07:55:27Z

@3pointer: The following test failed, say /retest to rerun all failed tests or /retest-required to rerun all mandatory failed tests:

Test name	Commit	Details	Required	Rerun command
tiprow_fast_test	`498160f`	link	true	`/test tiprow_fast_test`

Full PR test history. Your PR dashboard.

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/test-infra repository. I understand the commands that are listed here.

ti-chi-bot bot added release-note-none Denotes a PR that doesn't merit a release note. do-not-merge/invalid-title do-not-merge/work-in-progress Indicates that a PR should not merge because it is a work in progress. do-not-merge/needs-tests-checked labels Dec 28, 2023

ti-chi-bot bot added the size/XXL Denotes a PR that changes 1000+ lines, ignoring generated files. label Dec 28, 2023

3pointer changed the title ~~Per store concurrency~~ restore: use token bucket to balance download requests. Dec 28, 2023

ti-chi-bot bot removed the do-not-merge/invalid-title label Dec 28, 2023

ti-chi-bot bot added size/XL Denotes a PR that changes 500-999 lines, ignoring generated files. and removed size/XXL Denotes a PR that changes 1000+ lines, ignoring generated files. labels Jan 4, 2024

restore: use token bucket to balance download SSTs

d7d32b7

3pointer force-pushed the perStoreConcurrency branch from 2c60f61 to d7d32b7 Compare January 4, 2024 07:19

3pointer added 2 commits January 4, 2024 15:43

update config logic

5e77369

update

333fb78

3pointer marked this pull request as ready for review January 4, 2024 08:09

ti-chi-bot bot removed the do-not-merge/work-in-progress Indicates that a PR should not merge because it is a work in progress. label Jan 4, 2024

3pointer requested a review from Leavrth January 4, 2024 08:09

3pointer added 3 commits January 4, 2024 17:18

update granlurity control logic

70f3bbf

lint

71ae2a1

lint

4bc2b0d

YuJuncen reviewed Jan 5, 2024

View reviewed changes

3pointer and others added 3 commits January 5, 2024 16:03

Apply suggestions from code review

cac1281

Co-authored-by: 山岚 <[email protected]>

Merge branch 'perStoreConcurrency' of https://github.com/3pointer/tidb …

1faf7cb

…into patch_per_store_concurrency

address comments

ba75968

YuJuncen approved these changes Jan 5, 2024

View reviewed changes

ti-chi-bot bot added needs-1-more-lgtm Indicates a PR needs 1 more LGTM. approved labels Jan 5, 2024

3pointer added 4 commits January 5, 2024 17:03

address comment

8815582

Merge branch 'master' into patch_per_store_concurrency

bb4ee99

fix bazel

d088afa

update

1b63aea

Leavrth reviewed Jan 8, 2024

View reviewed changes

3pointer added 2 commits January 8, 2024 15:15

address comment

ac95cdd

update

498160f

Leavrth approved these changes Jan 8, 2024

View reviewed changes

br/pkg/restore/import.go Outdated Show resolved Hide resolved

ti-chi-bot bot added lgtm and removed needs-1-more-lgtm Indicates a PR needs 1 more LGTM. labels Jan 8, 2024

ti-chi-bot bot merged commit 5cc32b6 into pingcap:master Jan 8, 2024
22 of 23 checks passed

3pointer mentioned this pull request Jan 24, 2024

Tracking issues of restore performance improvement. #50701

Open

26 tasks

		rc.workerPool = utils.NewWorkerPool(c, "file")
		rc.concurrencyPerStore = c

restore: use token bucket to balance download requests. #49887

restore: use token bucket to balance download requests. #49887

Conversation

3pointer commented Dec 28, 2023 • edited Loading

What problem does this PR solve?

What changed and how does it work?

Check List

Release note

ti-chi-bot bot commented Dec 28, 2023

sre-bot commented Dec 28, 2023 • edited Loading

tiprow bot commented Dec 28, 2023

3pointer commented Dec 28, 2023

3pointer commented Jan 4, 2024

codecov bot commented Jan 4, 2024 • edited Loading

Codecov Report

3pointer commented Jan 4, 2024

ti-chi-bot bot commented Jan 4, 2024

tiprow bot commented Jan 4, 2024

3pointer commented Jan 4, 2024

tiprow bot commented Jan 4, 2024

YuJuncen left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

YuJuncen Jan 5, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

YuJuncen left a comment

Choose a reason for hiding this comment

YuJuncen Jan 5, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

3pointer commented Jan 8, 2024

3pointer commented Jan 8, 2024

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ti-chi-bot bot commented Jan 8, 2024

ti-chi-bot bot commented Jan 8, 2024

[LGTM Timeline notifier]

tiprow bot commented Jan 8, 2024

3pointer commented Dec 28, 2023 •

edited

Loading

sre-bot commented Dec 28, 2023 •

edited

Loading

codecov bot commented Jan 4, 2024 •

edited

Loading

YuJuncen Jan 5, 2024 •

edited

Loading

YuJuncen Jan 5, 2024 •

edited

Loading