jedarden/miroir
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

refactor(anti-entropy): remove dead with_mode_a_scaling and stale disabled-test comment (bf-no1g0)

Browse source 
The NOTE/TODO comment in tests/p13_8_anti_entropy.rs claimed a test
acceptance_4_mode_a_shard_partitioning existed-but-was-disabled due to an
API mismatch (with_mode_a_scaling). No such test lives in that file (it
jumps acceptance_3 -> bucket_isolation), and Mode A partitioning coverage
already lives inline in src/anti_entropy.rs (test_mode_a_anti_entropy_partitioning
plus the rendezvous minimal-reshuffling tests). Replaced the 4-line comment
with a one-line pointer to where the coverage actually lives.

Investigation also surfaced that with_mode_a_scaling itself is dead code:
the method is never called anywhere in the repo, and the fields it writes
(replica_group_id, num_pods) are write-only — never read. The live Mode A
scaling path routes through ModeACoordinator (with_mode_a_coordinator, read
in the run loop). Removed the method and its two orphaned fields.

Verified: grep for with_mode_a_scaling|acceptance_4_mode_a_shard_partitioning
under crates/ is empty; cargo build -p miroir-core succeeds; the external
p13_8_anti_entropy acceptance tests (8) pass.

Co-Authored-By: Claude <noreply@anthropic.com>
This commit is contained in:
jedarden 2026-07-10 16:33:32 -04:00
parent 0ade87a3de
commit a39b0e9df5
2 changed files with 1 additions and 24 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

20
crates/miroir-core/src/rebalancer_worker/anti_entropy_worker.rs
View file

@ -468,10 +468,6 @@ pub struct AntiEntropyWorker {
mode_a_coordinator: Option<Arc<ModeACoordinator>>,
/// Total shards in the cluster (for Mode A scaling).
_total_shards: u32,
/// This pod's replica group ID (for Mode A scaling).
replica_group_id: Option<u32>,
/// Total number of pods in Mode A scaling.
num_pods: Option<u32>,
/// RF (replication factor) for Mode A scaling.
_rf: usize,
/// Whether TTL is enabled for expired document handling (plan §13.14 interaction).
@ -520,8 +516,6 @@ impl AntiEntropyWorker {
pod_id,
mode_a_coordinator: None,
_total_shards: 0, // Will be set when Mode A is enabled
replica_group_id: None,
num_pods: None,
_rf: 2, // Default RF
ttl_enabled: false,
metrics_shards_scanned: None,
@ -537,20 +531,6 @@ impl AntiEntropyWorker {
self
}
/// Set Mode A scaling parameters (plan §14.6).
///
/// When enabled, each pod fingerprints and repairs only its rendezvous-owned shards.
///
/// # Parameters
///
/// - `replica_group_id`: This pod's ID in the pod pool (0-indexed)
/// - `num_pods`: Total number of pods running anti-entropy
pub fn with_mode_a_scaling(mut self, replica_group_id: u32, num_pods: u32) -> Self {
self.replica_group_id = Some(replica_group_id);
self.num_pods = Some(num_pods);
self
}
/// Set metrics callbacks.
pub fn with_metrics(
mut self,

5
crates/miroir-core/tests/p13_8_anti_entropy.rs
View file

@ -238,10 +238,7 @@ async fn test_acceptance_3_cdc_suppression() {
);
}
// NOTE: Test acceptance_4_mode_a_shard_partitioning is disabled because it requires
// the peer-discovery feature and ModeACoordinator. The test uses an old API
// (with_mode_a_scaling) that doesn't match the current implementation.
// TODO: Update this test to use ModeACoordinator or move it to a feature-gated module.
// Mode A shard-partitioning coverage (acceptance_4) lives inline in src/anti_entropy.rs.
/// Test that bucket-based diff isolates divergence to ~1/256 of PK space.
#[test]