Skip to content

BFT Consensus

Overview

The Byzantine Fault-Tolerant (BFT) consensus mechanism runs 3-phase PBFT when blocks are finalized. It requires n ≥ 3f + 1 nodes to tolerate f faulty or malicious nodes. This replaces the finalize_block() step in Event Submission for BFT mode.

For PoA and PoF consensus flows, see Consensus Mechanisms.

Requirement: Minimum 4 nodes to tolerate 1 Byzantine failure (n=4, f=1: 3×1+1=4).

Flow Diagram — 3-Phase PBFT

sequenceDiagram
    autonumber
    participant OS as ⚙️ OrderingService
    participant L as 👑 Leader Node
    participant V1 as 🖥️ Validator 1
    participant V2 as 🖥️ Validator 2
    participant Vf as 🖥️ Validator f

    OS->>L: Events batch ready → trigger consensus

    rect rgb(0, 0, 0, 0)
        Note over L,Vf: PHASE 1 — PRE-PREPARE
        L->>L: Assign sequence number, create PRE-PREPARE message
        L->>V1: PRE-PREPARE(view, seq, block_digest)
        L->>V2: PRE-PREPARE(view, seq, block_digest)
        L->>Vf: PRE-PREPARE(view, seq, block_digest)
    end

    rect rgb(0, 0, 0, 0)
        Note over L,Vf: PHASE 2 — PREPARE
        V1->>V1: Validate PRE-PREPARE, broadcast PREPARE
        V1->>L: PREPARE(view, seq, digest)
        V1->>V2: PREPARE(view, seq, digest)
        V2->>L: PREPARE(view, seq, digest)
        V2->>V1: PREPARE(view, seq, digest)
        Note over L: Collect 2f PREPARE votes
    end

    rect rgb(0, 0, 0, 0)
        Note over L,Vf: PHASE 3 — COMMIT
        L->>V1: COMMIT(view, seq, digest)
        L->>V2: COMMIT(view, seq, digest)
        V1->>L: COMMIT(view, seq, digest)
        V2->>L: COMMIT(view, seq, digest)
        Note over L: Collect 2f+1 COMMIT votes → finalize
        L->>L: Finalize & sign block
        L->>OS: Block committed → push to commit_queue
    end

Flow Diagram — View Change (Leader Failure)

sequenceDiagram
    autonumber
    participant V1 as 🖥️ Validator 1
    participant VM as 🔄 BFTViewChangeManager
    participant NEW as 👑 New Leader

    Note over V1: Leader timeout detected (no PRE-PREPARE received)

    V1->>VM: trigger_view_change(current_view, failed_leader)
    VM->>VM: view += 1
    VM->>VM: Broadcast VIEW-CHANGE to all validators
    VM->>VM: Collect f+1 VIEW-CHANGE votes
    VM->>NEW: Elect new leader: Validators[new_view % n]
    NEW->>NEW: Broadcast NEW-VIEW message
    NEW->>NEW: Restart from Phase 1 — PRE-PREPARE

Step-by-Step Breakdown

Step Description
PRE-PREPARE Leader assigns sequence number and broadcasts block digest to all validators
PREPARE Each validator validates the PRE-PREPARE, then broadcasts its own PREPARE vote. Leader collects 2f votes
COMMIT Leader broadcasts COMMIT. Each node collects 2f+1 COMMIT votes, then finalizes locally
View Change If leader silent > timeout: validators increment view, elect Validators[view % n] as new leader

Consensus Comparison

Algorithm Mechanism Fault Tolerance Use Case
PoA Identity-based, authority signs Validator reputation Private / internal networks
PoF Rotating leader, quorum height % n Distributed trust Consortium / multi-org
BFT 3-phase PBFT Up to f Byzantine nodes in 3f+1 Critical / adversarial environments

Error Handling

Condition Behavior
Leader timeout View Change triggered, new leader elected (Validators[new_view % n])
Validator sends invalid digest Vote discarded, not counted toward quorum
Network partition < f nodes Protocol continues if quorum (2f+1) still reachable
Network partition ≥ f+1 nodes Protocol halts until partition heals (safety over liveness)

Key Classes & Methods

Step Class / Method File
3-Phase PBFT BFTConsensus.run_consensus() consensus/bft/consensus.py
PRE-PREPARE BFTConsensus._send_pre_prepare() consensus/bft/consensus.py
PREPARE collect BFTConsensus._handle_prepare() consensus/bft/consensus.py
COMMIT finalize BFTConsensus._handle_commit() consensus/bft/consensus.py
View Change BFTViewChangeManager.trigger_view_change() consensus/bft/consensus.py
Transport ZmqTransport.send() / receive() network/zmq_transport.py