Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

kvclient(ticdc): fix kvclient takes too long time to recover (#3612) #3662

Merged
merged 5 commits into from
Dec 20, 2021
Merged

kvclient(ticdc): fix kvclient takes too long time to recover (#3612) #3662

merged 5 commits into from
Dec 20, 2021

Conversation

ti-chi-bot
Copy link
Member

This is an automated cherry-pick of #3612

close #3191
close flaky test in kvclient: #2694 #3302 #2349 #2688 #2747

What problem does this PR solve?

  1. When a tikv node which has 10k region leaders fail, we found it need more than 30min to recover.
  2. We think it is abnormal for the reason that pd and tikv only need about 30s to tag a node 'disconnect' and elect a new leader.

What is changed and how it works?

  1. Decrease the retry num when a new stream fail to establish to make other region try ASAP.
  2. Remove PartialClone when the region fail.

Reason

  • Expectation
    We use a tikv node which has about 3k region leaders as our comparison object. After test in normal case and abnormal case, we got follow results:
(1) ~ 1min for 3k region to finish init scan and get event (normal case)
(2) > 20min for all 3k region to recover the stream (tikv node fail)

But, pd and tikv only need about 30s to tag a node 'disconnect' and elect a new leader, so a reasonable timespan to recover is about:

       1min30s(normal case + region failover) ~ 5min (retry control in cdc)
  • Consider the probability of network jitter >> tikv temporary down >> tikv permanent down, we still need some retry logic, so
Decrease the retry num when a new stream fail to establish  to make other region try ASAP.
  • When a region fail and call RegionCache.OnRegionFail, it will (1) mark the store as 'needcheck' and fetch new store state of the store asynchronous, (2) move its leader to other peer to try next time and fetch new leader back. This mechanism is suitable in our case, but we find all failed region still connect to the old store which is quite strange. So we
Remove wrong PartialClone logic when the region fail in region_worker.

Result

3min30s to recover from 3k fail (after this pr)  compared to >20min (before)

Check List

Tests

  • Manual test (add detailed scripts or steps below)

Related changes

  • Need to cherry-pick to the release branch
  • Need to update the documentation
  • Need to update key monitor metrics in both TiCDC document and official document

Release note 

Please add a release note.
fix kvclient takes too long time to recover

@ti-chi-bot
Copy link
Member Author

[REVIEW NOTIFICATION]

This pull request has not been approved.

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.

@ti-chi-bot ti-chi-bot mentioned this pull request Nov 29, 2021
Merged
@ti-chi-bot ti-chi-bot added release-note Denotes a PR that will be considered when it comes time to generate release notes. do-not-merge/cherry-pick-not-approved component/kv-client TiKV kv log client component. size/M Denotes a PR that changes 30-99 lines, ignoring generated files. status/LGT2 Indicates that a PR has LGTM 2. type/cherry-pick-for-release-5.2 This PR is cherry-picked to release-5.2 from a source PR. labels Nov 29, 2021
@codecov-commenter
Copy link

Codecov Report

❗ No coverage uploaded for pull request base (release-5.2@636c0f5). Click here to learn what that means.
The diff coverage is n/a.

@@               Coverage Diff                @@
##             release-5.2      #3662   +/-   ##
================================================
  Coverage               ?   54.0232%           
================================================
  Files                  ?        193           
  Lines                  ?      21587           
  Branches               ?          0           
================================================
  Hits                   ?      11662           
  Misses                 ?       8829           
  Partials               ?       1096

@nongfushanquan nongfushanquan added cherry-pick-approved Cherry pick PR approved by release team. and removed do-not-merge/cherry-pick-not-approved labels Dec 16, 2021
@overvenus
Copy link
Member

/merge

@ti-chi-bot
Copy link
Member Author

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

Commit hash: eff9bef

@ti-chi-bot ti-chi-bot added the status/can-merge Indicates a PR has been approved by a committer. label Dec 20, 2021
@ti-chi-bot
Copy link
Member Author

@ti-chi-bot: Your PR was out of date, I have automatically updated it for you.

At the same time I will also trigger all tests for you:

/run-all-tests

If the CI test fails, you just re-trigger the test that failed and the bot will merge the PR for you after the CI passes.

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 ti-community-infra/tichi repository.

@ti-chi-bot ti-chi-bot merged commit 62101ec into pingcap:release-5.2 Dec 20, 2021
3AceShowHand pushed a commit to ti-chi-bot/tiflow that referenced this pull request Jan 13, 2022
@ti-chi-bot @3AceShowHand
kvclient(ticdc): fix kvclient takes too long time to recover (pingcap…
d93f48e 
…#3612) (pingcap#3662)
overvenus pushed a commit that referenced this pull request Jan 18, 2022
@ti-chi-bot @zhaoxinyu
@amyangfei @asddongmen @3AceShowHand
metrics(cdc): fix mq sink write row count metrics. (#4192) (#4322)
a5ce862 
* fix the txn_batch_size metric inaccuracy bug when the sink target is MQ

* address comments

* add comments for exported functions

* fix the compiling problem

* workerpool: limit the rate to output deadlock warning (#3775) (#3799)

* metrics: add changefeed checkepoint catch-up ETA (#3300) (#3311)

* pkg,cdc: do not use log package (#3902) (#3939)

* *: rename repo from pingcap/ticdc to pingcap/tiflow (#3957)

* kvclient(ticdc): fix kvclient takes too long time to recover (#3612) (#3662)

* tz (ticdc): fix timezone error (#3887) (#3910)

* http_*: add log for http api and refine the err handle logic (#2997) (#3306)

* ticdc/alert: add no owner alert rule (#3809) (#3831)

* etcd_worker: batch etcd patch (#3277) (#3393)

* ticdc/owner: Fix ddl special comment syntax error (#3845) (#3977)

* Clean old owner and old processor in release 5.2 branch (#4019)

* tests(ticdc): set up the sync diff output directory correctly (#3725) (#3745)

* owner,scheduler(cdc): fix nil pointer panic in owner scheduler (#2980) (#4007) (#4015)

* config(ticdc): Fix old value configuration check for maxwell protocol (#3747) (#3782)

* ticdc/processor: Fix backoff base delay misconfiguration (#3992) (#4027)

* sink(ticdc): cherry pick sink bug fix to release 5.2 (#4083) (#4119)

* metrics(ticdc): add resolved ts and add changefeed to dataflow (#4038) (#4103)

* kv(ticdc): reduce eventfeed rate limited log (#4072) (#4110)

close #4006

* This is an automated cherry-pick of #4192

Signed-off-by: ti-chi-bot <[email protected]>

* http_api (ticdc): fix http api 'get processor' panic. (#4117) (#4122)

close #3840

* cdc/sink: adjust kafka initialization logic (#3192) (#3568)

* This is an automated cherry-pick of #3192

Signed-off-by: ti-chi-bot <[email protected]>

* fix conflicts.

* This is an automated cherry-pick of #3682

Signed-off-by: ti-chi-bot <[email protected]>

* fix import.

* fix failpoint path.

* try to fix initialize.

* remove table_sink.

* remove initialization.

* remove owner.

* fix mq.

Co-authored-by: Ling Jin <[email protected]>
Co-authored-by: 3AceShowHand <[email protected]>

* sink (ticdc): fix a deadlock due to checkpointTs fall back in sinkNode (#4084) (#4098)

close #4055

* This is an automated cherry-pick of #4192

Signed-off-by: ti-chi-bot <[email protected]>

* fix conflicts.

* fix conflicts.

* fix conflicts.

Co-authored-by: zhaoxinyu <[email protected]>
Co-authored-by: amyangfei <[email protected]>
Co-authored-by: dongmen <[email protected]>
Co-authored-by: Ling Jin <[email protected]>
Co-authored-by: 3AceShowHand <[email protected]>
@purelind purelind added this to the v5.2.4 milestone Apr 18, 2022
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers
No reviews
Assignees

@maxshuang maxshuang

Labels
cherry-pick-approved Cherry pick PR approved by release team. component/kv-client TiKV kv log client component. release-note Denotes a PR that will be considered when it comes time to generate release notes. size/M Denotes a PR that changes 30-99 lines, ignoring generated files. status/can-merge Indicates a PR has been approved by a committer. status/LGT2 Indicates that a PR has LGTM 2. type/cherry-pick-for-release-5.2 This PR is cherry-picked to release-5.2 from a source PR.
Projects
None yet
Milestone
v5.2.4
Development

Successfully merging this pull request may close these issues.
6 participants
@ti-chi-bot @codecov-commenter @overvenus @purelind @maxshuang @nongfushanquan