Update litestream configuration

[epompeii]

This changeset continues the litestream configuration updates following:

1. Upgrade to Litestream 5.x #719
2. Add Litestream checkpoint configuration settings #727

[github-actions]

🤖 Claude Code Review

PR: #729
Base: devel
Head: u/ep/litestream-config
Commit: be52826c0bee60ba085272ffd86d9223cd1b7e91

---

Good — no remaining references to the old field names. Here's the review:

---

PR Review: Restructure Litestream snapshot/validation config into nested objects

Summary

This PR refactors the JsonLitestream config from flat fields (snapshot_interval, retention, validation_interval) into nested structs (JsonSnapshot, JsonValidation), mirroring the existing pattern established by JsonCheckpoint. Two commits: json_snapshot and from.

Positives

• Good structural alignment: Grouping interval and retention under JsonSnapshot (and interval under JsonValidation) mirrors how JsonCheckpoint is already structured. This is more intuitive.
• Proper destructuring: From impls and into_yaml use let Self { .. } = self; pattern, per project conventions.
• Thorough test updates: All 5 YAML generation tests are updated. The snapshot+validation test verifies the correct YAML output.
• Docs and OpenAPI spec updated: All 9 locale docs and openapi.json are consistent with the Rust changes.
• No stale references: No remaining uses of the old flat field names anywhere in the codebase.

Issues

1. Breaking API change — no migration path (Medium)

This is a breaking change to the server config schema. Existing bencher.json configs using the old flat fields (snapshot_interval, retention, validation_interval) will silently ignore those fields after upgrading — serde will skip unknown fields by default. Users would lose their snapshot/validation settings without any error or warning.

Suggestion: Consider adding #[serde(alias = "...")] or a custom deserializer that maps old field names, or at minimum document the breaking change in release notes.

2. Docs removed retention_check_interval (Low, pre-existing)

The old docs listed retention_check_interval as a configurable option, but it was never in the Rust struct. The new docs correctly remove it. This is fine — just noting it was a pre-existing doc/code mismatch that's now cleaned up.

3. JsonSnapshot with all-None fields serializes as empty object (Nit)

A JsonSnapshot { interval: None, retention: None } will serialize to snapshot: {} in YAML. The old code would produce None (no snapshot section) in this case. This is a minor behavioral difference that's unlikely to matter in practice — Litestream should handle an empty snapshot section gracefully — but worth noting.

4. Commit messages are terse

The commit messages from and json_snapshot don't describe what changed or why. Consider more descriptive messages for reviewability (e.g., "Restructure Litestream snapshot/validation config into nested objects").

Code Quality

• Follows project conventions (#[cfg_attr(feature = "schema", derive(JsonSchema))], skip_serializing_if, destructuring)
• No lint suppressions added
• No security concerns — this is config struct reshaping only
• No performance implications

Verdict

Clean refactoring that improves config ergonomics. The main concern is silent breakage for existing users (issue #1). If this is a Plus-only server config that's not widely deployed, the risk is lower, but a serde compatibility shim or migration note would be prudent.

---

_{Model: claude-opus-4-6}

[github-actions]

Bencher Report

Branch	u/ep/litestream-config
Testbed	ubuntu-22.04

Click to view all benchmark results

Benchmark	Latency	Benchmark Result microseconds (µs) (Result Δ%)	Upper Boundary microseconds (µs) (Limit %)
Adapter::Json	📈 view plot 🚷 view threshold	3.78 µs (+5.79%) Baseline: 3.57 µs	4.69 µs (80.62%)
Adapter::Magic (JSON)	📈 view plot 🚷 view threshold	3.86 µs (+9.16%) Baseline: 3.54 µs	4.62 µs (83.56%)
Adapter::Magic (Rust)	📈 view plot 🚷 view threshold	25.03 µs (-1.58%) Baseline: 25.43 µs	31.67 µs (79.04%)
Adapter::Rust	📈 view plot 🚷 view threshold	2.83 µs (+1.30%) Baseline: 2.80 µs	3.22 µs (87.99%)
Adapter::RustBench	📈 view plot 🚷 view threshold	2.84 µs (+1.74%) Baseline: 2.80 µs	3.22 µs (88.29%)
head_version_insert/batch/10	📈 view plot 🚷 view threshold	107.56 µs (+6.96%) Baseline: 100.56 µs	118.11 µs (91.07%)
head_version_insert/batch/100	📈 view plot 🚷 view threshold	242.27 µs (+1.67%) Baseline: 238.28 µs	264.46 µs (91.61%)
head_version_insert/batch/255	📈 view plot 🚷 view threshold	471.03 µs (+1.68%) Baseline: 463.23 µs	498.01 µs (94.58%)
head_version_insert/batch/50	📈 view plot 🚷 view threshold	170.42 µs (+5.57%) Baseline: 161.43 µs	182.15 µs (93.56%)
threshold_query/join/10	📈 view plot 🚷 view threshold	150.03 µs (+3.14%) Baseline: 145.46 µs	168.90 µs (88.83%)
threshold_query/join/20	📈 view plot 🚷 view threshold	166.03 µs (+3.78%) Baseline: 159.98 µs	185.68 µs (89.42%)
threshold_query/join/5	📈 view plot 🚷 view threshold	145.84 µs (+6.01%) Baseline: 137.57 µs	159.68 µs (91.33%)
threshold_query/join/50	📈 view plot 🚷 view threshold	206.11 µs (+2.36%) Baseline: 201.35 µs	230.54 µs (89.40%)

🐰 View full continuous benchmarking report in Bencher