Skip to content

Storages: Shutdown the LocalIndexScheduler before shutting down PageStorage/DeltaMergeStore (#9712)#9728

Closed
ti-chi-bot wants to merge 1 commit intopingcap:release-8.5from
ti-chi-bot:cherry-pick-9712-to-release-8.5
Closed

Storages: Shutdown the LocalIndexScheduler before shutting down PageStorage/DeltaMergeStore (#9712)#9728
ti-chi-bot wants to merge 1 commit intopingcap:release-8.5from
ti-chi-bot:cherry-pick-9712-to-release-8.5

Conversation

@ti-chi-bot
Copy link
Member

@ti-chi-bot ti-chi-bot commented Dec 13, 2024

This is an automated cherry-pick of #9712

What problem does this PR solve?

Issue Number: close #9714

Problem Summary:

Checkout the logging in issue.

From "Invalid page id, entry not exist [page_id=451.26] [resolve_id=451.26]" and the StackTrace DB::DM::Segment::restoreSegment(std::__1::shared_ptr<DB::Logger> const&, DB::DM::DMContext&, unsigned long) dbms/src/Storages/DeltaMerge/Segment.cpp:427 we can know that something is wrong when restoring the StableValueSpace of Segment. After adding some debugging message, we know that the error comes from restoring table_id=451, segment_id=8822.

Then dump the BlobData from PageStorage "meta". We found that the last two records persisted on disk is invalid as a SegmentMetaInfo.

Before the error happen, we saw that the table_id=451, segment_id=8822 had applied a SegmentUpdateMeta task after DeltaMergeStore::shutdown. Segment::replaceStableMetaVersion will access to a already shutdown PageStorage. Causing the WriteBatch wrote broken data.

[2024/12/10 00:04:03.027 +08:00] [INFO] [Server.cpp:1589] ["Shutting down storages."] [thread_id=1]
...
-- called by `DeltaMergeStore::shutdown` on table_id=451
[2024/12/10 00:04:03.050 +08:00] [INFO] [LocalIndexerScheduler.cpp:170] ["Removed 0 tasks, keyspace_id=4294967295 table_id=451"] [thread_id=1]
[2024/12/10 00:04:03.050 +08:00] [INFO] [RegionTable.cpp:152] ["remove table from RegionTable success, keyspace=4294967295 table_id=451"] [thread_id=1]
...
[2024/12/10 00:04:03.065 +08:00] [INFO] [LocalIndexerScheduler.cpp:69] ["LocalIndexerScheduler is destroying. Waiting scheduler and tasks to finish..."] [thread_id=1]
-- the running task on table_id=451, segment_id=8822, file_id=13172 is not removed. And 
[2024/12/10 00:05:28.470 +08:00] [INFO] [DMFileV3IncrementWriter.cpp:104] ["Write incremental update for DMFile, local_path=/data1/gengliqi/tidb/data/tiflash-9000/data/t_451/stable/dmf_13172 dmfile_path=/data1/gengliqi/tidb/data/tiflash-9000/data/t_451/stable/dmf_13172 old_meta_version=0 new_meta_version=1"] [thread_id=403]
[2024/12/10 00:05:28.474 +08:00] [INFO] [DeltaMergeStore_InternalSegment.cpp:772] ["EnsureStableLocalIndex - Finish building index, dm_files=[dmf_13172(v=0)]"] [source="keyspace=4294967295 table_id=451 segmentEnsureStableLocalIndex source_segment=<segment_id=8822 epoch=4 range=[249948,499989)>"] [thread_id=403]
-- `Segment::replaceStableMetaVersion` will access to a already shutdown PageStorage. Causing the WriteBatch wrote broken data.
[2024/12/10 00:05:28.479 +08:00] [INFO] [DeltaMergeStore_InternalSegment.cpp:690] ["SegmentUpdateMeta - Finish, old_segment=<segment_id=8822 epoch=4 range=[249948,499989)> new_segment=<segment_id=8822 epoch=5 range=[249948,499989)>"] [source="keyspace=4294967295 table_id=451"] [thread_id=403]
-- `DeltaMergeStore::~DeltaMergeStore` is called
[2024/12/10 00:05:28.479 +08:00] [INFO] [DeltaMergeStore.cpp:372] ["Release DeltaMerge Store start"] [source="keyspace=4294967295 table_id=451"] [thread_id=403]
[2024/12/10 00:05:28.479 +08:00] [INFO] [DeltaMergeStore.cpp:376] ["Release DeltaMerge Store end"] [source="keyspace=4294967295 table_id=451"] [thread_id=403]
[2024/12/10 00:05:28.479 +08:00] [INFO] [LocalIndexerScheduler.cpp:85] ["LocalIndexerScheduler is destroyed"] [thread_id=1]

What is changed and how it works?

Storages: Shutdown the LocalIndexScheduler before shutting down PageStorage/DeltaMergeStore
* Add a method `LocalIndexerScheduler::shutdown()` and ensure the running task are all finished before shutting down the GlobalPageStorage in `ContextShared::shutdown()`.
  • Add a method LocalIndexerScheduler::shutdown() and ensure the running task are all finished before shutting down the GlobalPageStorage in ContextShared::shutdown().
  • Add error message about the segment_id when restoreSegment failed
  • Add debugging tool command for reading the BlobData from PageStorage instance

Check List

Tests

  • Unit test
  • Integration test
  • Manual test (add detailed scripts or steps below)
1. Deploy a cluster with vector data
2. Create a script to run add/drop vector index repeatly 
`alter table vector_bench_test drop index idx_emb_l2;`
`alter table vector_bench_test add vector index idx_emb_l2 ((VEC_L2_DISTANCE(embedding)));`
3. Create a script to restart tiflash repeatly
4. Check whether tiflash restart successfully
5. TiFlash restart successfully after running those 2 scripts for 7 hours
  • No code

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

Fix an issue that TiFlash may fail to restart after applying vector index build jobs

@JaySon-Huang @ti-chi-bot
Storages: Shutdown the LocalIndexScheduler before shutting down PageS…
26aa2bd 
…torage/DeltaMergeStore (pingcap#9712)

close pingcap#9714

Storages: Shutdown the LocalIndexScheduler before shutting down PageStorage/DeltaMergeStore
* Add a method `LocalIndexerScheduler::shutdown()` and ensure the running task are all finished before shutting down the GlobalPageStorage in `ContextShared::shutdown()`.

Signed-off-by: JaySon-Huang <tshent@qq.com>
@ti-chi-bot ti-chi-bot added release-note Denotes a PR that will be considered when it comes time to generate release notes. size/L Denotes a PR that changes 100-499 lines, ignoring generated files. type/cherry-pick-for-release-8.5 This PR is cherry-picked to release-8.5 from a source PR. labels Dec 13, 2024
@ti-chi-bot ti-chi-bot mentioned this pull request Dec 13, 2024
Merged
12 tasks
@ti-chi-bot ti-chi-bot bot added the cherry-pick-approved Cherry pick PR approved by release team. label Dec 13, 2024
@ti-chi-bot
Copy link
Contributor

ti-chi-bot bot commented Dec 13, 2024

[APPROVALNOTIFIER] This PR is NOT APPROVED

This pull-request has been approved by:
Once this PR has been reviewed and has the lgtm label, please assign hongyunyan for approval. For more information see the Code Review Process.

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

Details Needs approval from an approver in each of these files:

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 ti-chi-bot bot added size/S Denotes a PR that changes 10-29 lines, ignoring generated files. and removed do-not-merge/cherry-pick-not-approved size/L Denotes a PR that changes 100-499 lines, ignoring generated files. labels Dec 13, 2024
@JaySon-Huang
Copy link
Contributor

JaySon-Huang commented Dec 13, 2024 •
edited
Loading

/hold

The bug has been fixed in release-8.5 in a previous PR #9713. This PR pick some other code diffs. Merge it later.

@ti-chi-bot 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 Dec 13, 2024
@JaySon-Huang
Copy link
Contributor

JaySon-Huang commented Dec 13, 2024

/close

The unnecessary change does not affect the bug fix. The changes will be merged along with other PR.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

@JaySon-Huang JaySon-Huang

Labels

cherry-pick-approved Cherry pick PR approved by release team. do-not-merge/hold Indicates that a PR should not merge because someone has issued a /hold command. release-note Denotes a PR that will be considered when it comes time to generate release notes. size/S Denotes a PR that changes 10-29 lines, ignoring generated files. type/cherry-pick-for-release-8.5 This PR is cherry-picked to release-8.5 from a source PR.

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@ti-chi-bot @JaySon-Huang