SDSTOR-22293: handle logdev recovery corner case

[Besroy]

also add relevant knowledge to docs/structures to support AI in future triaging

[Besroy]

@JacksonYao287 Since there isn’t a clean, fully compatible way to handle this case right now (for example, the two sbs in logstore are not atomic), I only changed logstream recovery in this PR. PTAL.

[codecov-commenter]

⚠️ Please install the to ensure uploads and comments are reliably processed by Codecov.

Codecov Report

❌ Patch coverage is 70.00000% with 6 lines in your changes missing coverage. Please review.
⚠️ Please upload report for BASE (stable/v7.x@ef9ab9d). Learn more about missing BASE report.

Files with missing lines	Patch %	Lines
src/lib/logstore/log_dev.cpp	54.54%	0 Missing and 5 partials ⚠️
src/lib/device/journal_vdev.cpp	88.88%	0 Missing and 1 partial ⚠️
❗ Your organization needs to install the Codecov GitHub app to enable full functionality.

Additional details and impacted files

@@              Coverage Diff               @@
##             stable/v7.x     #894   +/-   ##
==============================================
  Coverage               ?   48.30%           
==============================================
  Files                  ?      110           
  Lines                  ?    12972           
  Branches               ?     6232           
==============================================
  Hits                   ?     6266           
  Misses                 ?     2564           
  Partials               ?     4142           

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[xiaoxichen]

TL;DR

Instead of fighting with the corner cases, it seems like we dont need the dynamic feature at all, we knows the journal vdev size and we know our #PG per SM, any possibility of static allocation / pre-fill works with minimal modification ?

[JacksonYao287]

it`s a big change, will review it later this week.

[Besroy]

TL;DR

Instead of fighting with the corner cases, it seems like we dont need the dynamic feature at all, we knows the journal vdev size and we know our #PG per SM, any possibility of static allocation / pre-fill works with minimal modification ?

I don't think static allocation/pre-fill would be a minimal modification. It would fundamentally change the chunk chain structure for each PG into a pre-allocated ring, requiring significant redesign across several areas:

• Truncation logic: instead of releasing chunks back to the pool, it becomes a ring-pointer advance
• Wrap-around handling: offset_to_chunk(), tail_offset(), and bounds checks all need circular semantics
• New edge cases to evaluate: what happens when data space is exhausted but the log hasn't caught up (data-full vs. log-full mismatch), and how to handle uneven request distribution across PGs (one PG's ring fills up while another sits idle)
• Superblock changes: data_start_offset would need to become a relative offset within the ring, and we'd need to decide between a fixed ring head vs. a dynamic head marker — either way, the on-disk format changes

If we want to handle this issue, this PR is a quick fix. If we want to revisit the chunk allocation model, that should be treated as a standalone architectural initiative — worth looping in all affected members to evaluate and plan for the right release.

[Besroy]

it`s a big change, will review it later this week.

Thanks! raft.md and raft_repl_dev_log_dev.md just provide related knowledge (future usage for AI triage more issues), you can skip them.

[JacksonYao287]

not sure do we need to update m_logdev_meta and persist with the correct offset in private data?