[ENH](wal3): add optional upload fault injector

[rescrv]

Description of changes

Add a generic fragment-manager shim in wal3 that can wrap
fragment uploaders with an optional fault injector.

Wire chroma-log-service's FaultRegistry through the S3,
repl, and copy/open paths so fragment uploads can be
delayed or failed via the wal3.fragment_upload label.

Keep the injector optional for callers that do not want
fault injection, and add focused wal3 and log-service
coverage for the new path.

Test plan

CI

Migration plan

N/A

Observability plan

This is about fault injection.

Documentation Changes

N/A

Co-authored-by: AI

[github-actions]

Reviewer Checklist

Please leverage this checklist to ensure your code review is thorough before approving

Testing, Bugs, Errors, Logs, Documentation

• Can you think of any use case in which the code does not behave as intended? Have they been tested?
• Can you think of any inputs or external events that could break the code? Is user input validated and safe? Have they been tested?
• If appropriate, are there adequate property based tests?
• If appropriate, are there adequate unit tests?
• Should any logging, debugging, tracing information be added or removed?
• Are error messages user-friendly?
• Have all documentation changes needed been made?
• Have all non-obvious changes been commented?

System Compatibility

• Are there any potential impacts on other parts of the system or backward compatibility?
• Does this change intersect with any items on our roadmap, and if so, is there a plan for fitting them together?

Quality

• Is this code of a unexpectedly high quality (Readability, Modularity, Intuitiveness)

[propel-code-bot]

wal3 optional fragment upload fault injection and log-service wiring

This PR adds a new optional fault-injection path for fragment uploads in wal3 and threads it through log-service so upload operations can be delayed or failed using the wal3.fragment_upload label. The core change introduces fault-aware wrappers around fragment manager factories and uploaders, while preserving default behavior when no injector is configured.

To support this, the FragmentManagerFactory contract is expanded to expose make_fragment_uploader and write_options, and concrete implementations in rust/wal3/src/interfaces/s3/mod.rs, rust/wal3/src/interfaces/repl/mod.rs, and test factories were updated accordingly. log-service adds feature-gated integration with FaultRegistry, passes the injector through open/copy/fork/update/GC paths, and adds focused tests validating failure and recovery when faults are injected.

This summary was automatically generated by @propel-code-bot

[propel-code-bot]

The PR is functionally solid, with minor recommendations to reduce duplicated factory logic and clarify public API visibility/docs for newly exposed fault-injection traits and types.

Status: Minor Suggestions | Risk: Low

Issues Identified & Suggestions

• Duplicate consumer construction logic; extract shared builder helpers. rust/wal3/src/interfaces/s3/mod.rs, rust/wal3/src/interfaces/repl/mod.rs
• Public #[doc(hidden)] APIs are confusing; make crate-private or document. rust/wal3/src/interfaces/mod.rs, rust/wal3/src/lib.rs

Review Details

📁 6 files reviewed | 💬 3 comments

👍 / 👎 individual comments to help improve reviews for you

[propel-code-bot]

[Maintainability] [CodeDuplication] The make_consumer implementation in FragmentManagerFactoryWithUploader for ReplicatedFragmentManagerFactory (lines 129–137) is identical to the make_consumer in FragmentManagerFactory for ReplicatedFragmentManagerFactory (lines ~108–114 in the after state): both call FragmentReader::new(self.repl.clone(), self.preferred, storages, Arc::clone(&self.read_repair_semaphore)). The same duplication exists in rust/wal3/src/interfaces/s3/mod.rs for S3FragmentPuller. Extract a build_fragment_reader helper (matching build_fragment_uploader) to eliminate this duplication and prevent drift.

Context for Agents

[CodeDuplication] The `make_consumer` implementation in `FragmentManagerFactoryWithUploader for ReplicatedFragmentManagerFactory` (lines 129–137) is identical to the `make_consumer` in `FragmentManagerFactory for ReplicatedFragmentManagerFactory` (lines ~108–114 in the after state): both call `FragmentReader::new(self.repl.clone(), self.preferred, storages, Arc::clone(&self.read_repair_semaphore))`. The same duplication exists in `rust/wal3/src/interfaces/s3/mod.rs` for `S3FragmentPuller`. Extract a `build_fragment_reader` helper (matching `build_fragment_uploader`) to eliminate this duplication and prevent drift.

File: rust/wal3/src/interfaces/repl/mod.rs
Line: 136

[propel-code-bot]

Review found no issues; the changes appear well-implemented and low risk.

Status: No Issues Found | Risk: Low

Review Details

📁 14 files reviewed | 💬 0 comments