fix(index): fine-tune fjall default for expected write throughput

[scarmuega]

Summary by CodeRabbit

• Chores
  • Updated default database configuration settings, including adjustments to transaction handling, memory allocation, worker thread concurrency, and cache behavior for improved consistency and performance.

[coderabbitai]

📝 Walkthrough

Walkthrough

This PR replaces the derived Default implementation for FjallIndexConfig with a manual one that establishes explicit default values for Fjall database configuration parameters including journal size, flush behavior, L0 threshold, worker threads, and memtable size.

Changes

Cohort / File(s)	Summary
Fjall Configuration Defaults crates/core/src/config.rs	Removed Default from derive attributes and added manual Default implementation with explicit values: max_journal_size=1024, flush_on_commit=false, l0_threshold=8, worker_threads=8, memtable_size_mb=128.

Estimated code review effort

🎯 2 (Simple) | ⏱️ ~10 minutes

Possibly related PRs

• txpipe/dolos#845: Directly modifies and propagates the same Fjall configuration fields (max_journal_size, flush_on_commit, l0_threshold, worker_threads, memtable_size_mb) across StateStore/IndexStore, making use of these newly-explicit defaults.

Poem

🐰 A config struct once left to chance,
Now defaults take their proper stance.
Journals set at 1024's might,
Threads and levels, tuned just right! 🎯

🚥 Pre-merge checks | ✅ 2 | ❌ 1

❌ Failed checks (1 warning)

Check name	Status	Explanation	Resolution
Docstring Coverage	⚠️ Warning	Docstring coverage is 0.00% which is insufficient. The required threshold is 80.00%.	Write docstrings for the functions missing them to satisfy the coverage threshold.

✅ Passed checks (2 passed)

Check name	Status	Explanation
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title 'fix(index): fine-tune fjall default for expected write throughput' directly and specifically describes the main change: adjusting FjallIndexConfig default values to optimize for write throughput.

_{✏️ Tip: You can configure your own custom pre-merge checks in the settings.}

✨ Finishing touches

• 📝 Generate docstrings

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Post copyable unit tests in a comment
• Commit unit tests in branch fix/fjall-defaults

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands and usage tips.}

[coderabbitai]

Caution

Some comments are outside the diff and can’t be posted inline due to platform limitations.

⚠️ Outside diff range comments (1)

crates/core/src/config.rs (1)

354-392: ⚠️ Potential issue | 🟠 Major

Per-field #[serde(default)] bypasses the new defaults during partial deserialization.

Each field carries #[serde(default)], which means serde uses Option::default() (i.e., None) for every omitted field — not the values from your manual Default impl. The custom defaults only take effect when the entire FjallIndexConfig is absent and IndexStoreConfig::default() is called. If a user supplies a partial config (e.g., only path), the tuned values for max_journal_size, worker_threads, etc. will silently be None.

Replace the per-field #[serde(default)] annotations with a single struct-level #[serde(default)] so that missing fields fall back to the Default impl:

Proposed fix

-#[derive(Serialize, Deserialize, Clone, Debug)]
+#[derive(Serialize, Deserialize, Clone, Debug)]
+#[serde(default)]
 pub struct FjallIndexConfig {
     /// Optional path override. If relative, resolved from storage root.
     /// If not specified, defaults to `<storage.path>/index`.
-    #[serde(default)]
     pub path: Option<PathBuf>,
     /// Size (in MB) of memory allocated for caching.
-    #[serde(default)]
     pub cache: Option<usize>,
     /// Maximum journal size in MB (default: 1024).
-    #[serde(default)]
     pub max_journal_size: Option<usize>,
     /// Flush journal after each commit (default: false).
-    #[serde(default)]
     pub flush_on_commit: Option<bool>,
     /// L0 compaction threshold (default: 8, lower = more aggressive).
-    #[serde(default)]
     pub l0_threshold: Option<u8>,
     /// Number of background compaction worker threads (default: 8).
-    #[serde(default)]
     pub worker_threads: Option<usize>,
     /// Memtable size in MB before flush (default: 128).
-    #[serde(default)]
     pub memtable_size_mb: Option<usize>,
 }

🧹 Nitpick comments (1)

crates/core/src/config.rs (1)

209-236: Consider applying similar tuned defaults to FjallStateConfig.

FjallStateConfig has the same set of tunable fields (max_journal_size, l0_threshold, worker_threads, memtable_size_mb, etc.) but still derives Default, so all fields default to None. The doc comments even reference non-None defaults (e.g., "default: 4" for l0_threshold, "default: 64" for memtable_size_mb) that don't match the derived behavior. If the index store benefits from explicit defaults for write throughput, the state store likely does too.