Course contentsShow
← Computer Science
Lesson 2673 of 2,87259. Consensus and ReplicationFree lesson

Paxos Roles: Proposers, Acceptors, and Learners

Learning the three distinct roles in Paxos and how they interact to achieve consensus in a distributed system.

Paxos Roles: Proposers, Acceptors, and Learners

You'll learn how Paxos divides consensus work among three specialized roles that collaborate to agree on a single value in a distributed system.

The Three Roles

Unlike Raft's single-role servers (follower/candidate/leader), Paxos separates consensus responsibilities into three distinct roles that can be played by different or overlapping nodes:

Proposers

Proposers initiate the consensus process by suggesting values. Think of them as legislators proposing bills. A proposer:

  • Generates unique proposal numbers
  • Sends proposals to acceptors
  • Drives the two-phase protocol (prepare and accept)

Multiple proposers can exist simultaneously, competing to get their values chosen.

Acceptors

Acceptors form the voting body that decides which value becomes consensus. Like a parliament, they:

  • Receive proposals from proposers
  • Vote on proposals according to Paxos rules
  • Store promises and accepted values
  • Ensure safety by only accepting proposals that respect previous promises

A majority quorum of acceptors must agree for consensus to be reached—this guarantees safety even under failures.

Learners

Learners discover what value was chosen without participating in voting. They:

  • Passively observe accepted values
  • Learn the consensus result once a quorum has accepted
  • Apply the chosen value to their state machines

Learners are like citizens who learn which bill became law without voting themselves.

How They Interact

The typical flow: A proposer sends a proposal → acceptors vote using majority quorums → learners discover the chosen value. Nodes can play multiple roles simultaneously (e.g., a server might be both an acceptor and learner).

Key Takeaway: Paxos achieves consensus through role specialization—proposers suggest values, acceptors vote in quorums to ensure safety, and learners passively discover the result. This separation enables flexible deployment and fault tolerance.