Milestone 8 (phases 1-4): - Shard-aware WAL segment naming, BatchHeader v2, ShardRouter - Transport trait, InProcessTransport, WalShipper, FollowerDb - HLC, PNCounter, LWWRegister, CrdtSignalState, ReconciliationEngine - Session replication bridge with SeqNo/HWM, idempotency store Forage application: - Multi-source discovery engine with MAB exploration - Embedding-based label system, server handlers, UI refresh Other: - QUICKSTART.md, README.md, milestone-8 planning docs - Hard negative union semantics, RLHF export enhancements - Recovery benchmark and visibility test expansions - Split 8 oversized source files per CODING_GUIDELINES §9 Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
3.2 KiB
3.2 KiB
Task 07: Replication Integration Tests
Delivers
Integration tests in tidal/tests/m8p2_replication.rs covering the full replication stack: leader->follower segment delivery, decay score equivalence to 6 decimal places, follower read-only enforcement, lag convergence, and segment corruption rejection.
Complexity: M
Dependencies
- Tasks 01-06 complete
Technical Design
// tidal/tests/m8p2_replication.rs
use tidaldb::{TidalDb, TidalDbBuilder, NodeRole, ShardId, RegionId, NodeConfig};
use tidaldb::replication::{InProcessTransportFactory, ReplicationLagGauge};
fn leader_config(data_dir: &Path) -> Config {
Config {
cluster: NodeConfig {
role: NodeRole::Leader,
shard_id: ShardId(0),
..Default::default()
},
..Config::with_data_dir(data_dir)
}
}
fn follower_config(data_dir: &Path) -> Config {
Config {
cluster: NodeConfig {
role: NodeRole::Follower,
shard_id: ShardId(0),
..Default::default()
},
..Config::with_data_dir(data_dir)
}
}
#[tokio::test]
async fn replication_decay_scores_match() {
// Leader writes 1,000 signals.
// Follower replays all segments.
// Verify: read_decay_score on follower matches leader to 6 decimal places.
}
#[tokio::test]
async fn follower_rejects_writes() {
// Open follower. Attempt signal() write.
// Verify: returns TidalError::ReadOnly.
}
#[tokio::test]
async fn follower_serves_retrieve_queries() {
// Leader writes items + signals.
// Follower applies.
// Follower.retrieve() returns ranked results.
}
#[tokio::test]
async fn replication_lag_converges_to_zero() {
// Leader writes 500 segments.
// Wait for follower to apply all.
// Assert: lag_seqno(follower) == 0 within 5 seconds.
}
#[tokio::test]
async fn corrupted_segment_is_rejected() {
// Manually corrupt BLAKE3 checksum in segment bytes.
// Send to follower via transport.
// Verify: segment is not applied (decay scores unchanged).
}
#[tokio::test]
async fn leader_restart_follower_continues() {
// Leader writes 100 signals.
// Leader shuts down.
// Follower serves read queries from replayed state.
// Leader restarts; ships remaining segments.
// Follower catches up.
}
#[tokio::test]
async fn idempotent_segment_replay() {
// Ship same segment twice to follower.
// Verify: signal counts NOT doubled (seqno idempotency).
}
Acceptance Criteria
- All 7 integration tests pass under
cargo test --test m8p2_replication - Test
replication_decay_scores_match: leader 1K signals -> follower matches to 6 decimal places - Test
follower_rejects_writes:TidalError::ReadOnlyon all write methods - Test
follower_serves_retrieve_queries: follower returns correct ranked results - Test
replication_lag_converges_to_zero: lag = 0 within 5 seconds of leader quiesce - Test
corrupted_segment_is_rejected: corrupt checksums rejected, no state change - Test
leader_restart_follower_continues: follower serves reads after leader crash - Test
idempotent_segment_replay: no double-counting on duplicate segments cargo clippy -D warningsandcargo fmtpass