rivven-cluster

Distributed cluster coordination for the Rivven event streaming platform.

Overview

rivven-cluster provides the distributed coordination layer for Rivven, including leader election, failure detection, and partition assignment.

Features

Feature	Description
Raft Consensus	Leader election and log replication using OpenRaft
redb Storage	Pure Rust persistent storage (zero C dependencies)
SWIM Gossip	Failure detection and membership management
Partitioning	Consistent hashing for partition assignment
QUIC Transport	Efficient, secure node-to-node communication

Why redb?

Rivven uses redb instead of RocksDB for Raft log storage:

Aspect	redb	RocksDB
Build time	~10s	2-5 min
Binary size	Minimal	+10-15 MB
Cross-compile	✅ Works everywhere	❌ Needs C++ toolchain
Docker musl	✅ Works	❌ Needs musl-g++
ACID	✅ Full	✅ Full

Architecture

┌─────────────────────────────────────────────────────────────────┐
│                        rivven-cluster                           │
├─────────────────────────────────────────────────────────────────┤
│  ┌──────────────┐    ┌──────────────┐    ┌──────────────┐       │
│  │    Raft      │    │    SWIM      │    │  Partition   │       │
│  │  Consensus   │    │   Gossip     │    │  Manager     │       │
│  └──────────────┘    └──────────────┘    └──────────────┘       │
│         │                   │                   │               │
│         └───────────────────┼───────────────────┘               │
│                             │                                   │
│                    ┌────────────────┐                           │
│                    │ QUIC Transport │                           │
│                    └────────────────┘                           │
└─────────────────────────────────────────────────────────────────┘

Components

Raft Consensus

Used for metadata replication and leader election:

use rivven_cluster::{RaftNode, RaftConfig};

let config = RaftConfig {
    node_id: 1,
    peers: vec![2, 3],
    election_timeout_ms: 150..300,
    heartbeat_interval_ms: 50,
};

let node = RaftNode::new(config).await?;

SWIM Gossip

Decentralized failure detection:

use rivven_cluster::{SwimNode, SwimConfig};

let config = SwimConfig {
    bind_addr: "0.0.0.0:7946".parse()?,
    known_peers: vec!["node2:7946".parse()?],
    protocol_period_ms: 1000,
    suspect_timeout_ms: 5000,
};

let node = SwimNode::new(config).await?;

Partition Assignment

Consistent hashing with virtual nodes:

use rivven_cluster::{PartitionManager, HashRing};

let ring = HashRing::new(128); // 128 virtual nodes per physical node
ring.add_node("node1")?;
ring.add_node("node2")?;

let owner = ring.get_partition_owner("my-topic", 0)?;

Configuration

cluster:
  node_id: 1
  bind_address: "0.0.0.0:9093"
  
  raft:
    peers:
      - "node2:9093"
      - "node3:9093"
    election_timeout_min_ms: 150
    election_timeout_max_ms: 300
    
  gossip:
    bind_port: 7946
    known_peers:
      - "node2:7946"
    protocol_period_ms: 1000
    
  partitioning:
    virtual_nodes: 128
    replication_factor: 3

Documentation

License

Apache-2.0. See LICENSE.

rivven-cluster 0.0.5