Struct MiniCluster 

pub struct MiniCluster {
    pub nodes: Vec<NodeHandle>,
    pub rules: Option<Arc<NetworkRules>>,
    pub snapshot: Option<Arc<AssignmentSnapshotStore>>,
}

Builder + wrapper for a set of in-process cluster nodes.

Fields

nodes: Vec<NodeHandle>

The nodes in ID order: nodes[0] has the lowest instance_id and is the leader under normal operation.

rules: Option<Arc<NetworkRules>>

Shared network rules. None for plain Self::spawn clusters (default UDP transport); Some when built via Self::spawn_partitionable.

snapshot: Option<Arc<AssignmentSnapshotStore>>

Shared assignment snapshot store handed to every node’s controller (so a snapshot written by one node is visible to all). None unless built via Self::spawn_with_snapshot.

Implementations

impl MiniCluster

pub async fn spawn(n: usize) -> Self

Spin up n nodes with the default UDP transport. See Self::spawn_partitionable for a variant that allows simulating network partitions.

Panics

Panics if any chitchat instance fails to start (UDP bind failure, usually indicates port contention).

pub async fn spawn_partitionable(n: usize) -> Self

Spin up n nodes with a PartitionableTransport wrapping the default UDP transport. The returned cluster carries Self::rules so tests can call NetworkRules::partition / NetworkRules::heal during the test.

Panics

Same as Self::spawn.

pub async fn spawn_with_snapshot( n: usize, snapshot: Arc<AssignmentSnapshotStore>, ) -> Self

Spin up n nodes sharing the given AssignmentSnapshotStore. Each node’s controller is constructed with the same store, so a snapshot written by any node is visible to every other node and survives across a full cluster restart (provided the object store does).

pub async fn join_node(&mut self, instance_id: NodeId)

Join one new node with the given instance_id into an already-running cluster. Seeds from nodes[0] for discovery. Useful for testing rejoin-after-kill and elastic scale-up scenarios.

Panics

Panics if the cluster is empty (nothing to seed from) or if chitchat fails to bind a loopback port.

pub fn addrs(&self) -> Vec<SocketAddr>

Parse and collect node gossip addresses — useful for building partition rules via NetworkRules::partition.

Panics

Panics if any node’s gossip_addr doesn’t parse as a SocketAddr (only possible if the MiniCluster was constructed with malformed input).

pub async fn wait_for_convergence( &self, deadline: Duration, ) -> Result<(), String>

Wait until every node sees every other node as a peer, or until deadline passes.

Errors

Returns Err(String) on timeout, describing which nodes haven’t converged yet, or if a discovery peers() call fails.

pub async fn shutdown(self)

Shut every node down cleanly.