PageStorage: Fix empty page cause TiFlash failed to start

[JaySon-Huang]

What problem does this PR solve?

Issue Number: close #9282

Problem Summary:

By dumping the entries from PageStorage WAL, we can locate that two entries cause this issue:

[PageDirectoryFactory.cpp:225] ["{type:VAR_ENT, page_id:0x020102000000009CA4430A08, ori_id:0x.0, version:108969363.0, entry:PageEntry{file: 551, offset: 0x15376, size: 0, checksum: 0x0, tag: 0, field_offsets: [], checkpoint_info: invalid}, being_ref_count:1}"] [thread_id=1]
[PageDirectoryFactory.cpp:225] ["{type:PUT    , page_id:0x020102000000018204A17C03, ori_id:0x.0, version:127726467.0, entry:PageEntry{file: 551, offset: 0x15373, size: 15, checksum: 0x12D46C1C0D5B51FE, tag: 0, field_offsets: [], checkpoint_info: invalid}, being_ref_count:1}"] [thread_id=1]

The first entry is a page with data.size == 0, placed at blob_id=551, offset=0x15376. And later another page is placed at blob_id=551, offset=0x15373 with data.size == 15.
The first page split the free block into smaller pieces free blocks.
The second page will cover the place of the first page. And our code can not handle this situation with "empty page data".

What is changed and how it works?

• For handling the existing empty pages, we won't let it split the free block into small pieces:
  • when STDMapSpaceMap::markUsedImpl accept a block of length == 0, it will directly return true instead of splitting the free block into smaller pieces.
• For newly created empty pages, we store them at the front of BlobFile to avoid further issues:
  • BlobStats::BlobStat::getPosFromStat and STDMapSpaceMap::searchInsertOffset accepts a block of length == 0, it will return insert_offset == 0, which we can "store the empty pages" at the front of the BlobFile for newly created empty pages.
• In BlobStore::read(...), we ensure that reading empty pages won't cause NPE or other unexpected exceptions

PageStorage: Fix empty page cause TiFlash failed to start

Check List

Tests

• Unit test
• Integration test
• Manual test (add detailed scripts or steps below)
• 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 write node may fail to restart under disaggregated arch

[JaySon-Huang]

We will test it later, now just keep the same behavior

[JaySon-Huang]

/retest

[purelind]

/retest

[CalvinNeo]

lgtm

[ti-chi-bot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: CalvinNeo, JinheLin

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 [CalvinNeo,JinheLin]

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:

• 2024-08-05 03:29:33.70887058 +0000 UTC m=+236303.575969669: ☑️ agreed by JinheLin.
• 2024-08-05 04:01:01.964005981 +0000 UTC m=+238191.831105069: ☑️ agreed by CalvinNeo.

[ti-chi-bot]

In response to a cherrypick label: new pull request created to branch release-8.1: #9285.

[ti-chi-bot]

In response to a cherrypick label: new pull request created to branch release-7.5: #9286.