Module cross_dc 

Cross-datacenter async replication for distributed vector indexes

This module implements asynchronous replication of vector index updates across multiple datacenters with configurable lag tolerance.

Design

Each datacenter hosts a DcReplicationNode. One DC is designated the primary (or leader); others are replicas. Index mutations are buffered in the primary and asynchronously shipped to replicas.

Key features:

• Configurable lag tolerance (max acceptable replication lag)
• Per-replica lag tracking and alerting
• Automatic catch-up on reconnection
• Conflict resolution strategies for network partitions
• Bandwidth throttling to avoid saturating WAN links

Consistency Model

Cross-DC replication is eventually consistent with configurable lag bounds. Reads on replicas may return stale data. For strong consistency, use the Raft cluster within a single DC.

Structs

ConflictRecord

Record of a detected conflict

CrossDcConfig

Configuration for cross-DC replication

CrossDcCoordinator

Cross-DC replication coordinator

CrossDcStats

Statistics for cross-DC replication

PrimaryDcManager

The primary datacenter replication manager

ReplicaDcManager

Replica datacenter replication receiver

ReplicaTracker

Per-replica tracking state (maintained by primary)

ReplicationEntry

A single replication entry

ReplicationHealth

Overall replication health summary

Enums

ConflictResolutionStrategy

Strategy for resolving write conflicts between DCs

ReplicaStatus

Status of a replica DC connection

ReplicationOperation

Types of replication operations

Type Aliases

DcId

Datacenter identifier

ReplicationSeq

Replication sequence number (monotonically increasing per primary)