Mult-Paxos

🏮 This repository contains the Implementation for the graduate lesson Concept and design of Distributed System (2022 Spring, advised by Zhi Yang) in Peking University.

Getting Start

• Setup the conda environment.

conda env create -f multi-paxos.yaml

• Autogen the Python code for grpc and protobuf

python -m grpc_tools.protoc -I ./ --python_out=./ --grpc_python_out=. ./phxkv.proto

If you finish running this command without error, you will see two new files phxkv_pb2_grpc.py and phxkv_pb2.py in the current directory.

Usage

We implement the Multi-Paxos Algorithm and a distributed database based on our Paxos consensus protocol.

Echo Usage

# Running the server
# Without master leases:
python server.py A|B|C
# With master leases:
python server.py --master A|B|C

# Running the client
python client.py A|B|C <value>

Paxos-based Distributed Database

• Setup the server A, B and C.

$ python kv_grpc_server_main.py --help
usage: kv_grpc_server_main.py [-h] [--uid {A,B,C}] [--master]

Paxos-based Distributed KV Store Server

optional arguments:
  -h, --help     show this help message and exit
  --uid {A,B,C}  UID of the server. Must be A, B, or C
  --master       If specified, a dedicated master will be used. If one server
                 specifies this flag, all must         

• Interact with the server.

$ python kv_grpc_client_cli.py
Welcome to the KV Client CLI.

Available commands:
  help                           - print this help message
  q                              - quit
  connect <UID>                  - connect the gRPC server by UID
  disconnect                     - disconnect current gRPC server
  get <key>                      - get the value of the key
  put <key> <value>              - set the value of the key
  delete <key>                   - delete the key

Available Servers:
  A: ('127.0.0.1', 50054)
  B: ('127.0.0.1', 50055)
  C: ('127.0.0.1', 50056)

Please select a server via the command `connect <UID>` first.

Content

• Basic multi-paxos
  • client.py implements the client.
  • server.py implements the server.
  • config.py contains required information, such as server IP, port, and path of state files.
  • messenger.py contains the message-passing strategy.
  • composable_paxos.py contains the core paxos algorithm, including the paxos instance (proposer, accepter, and learner) and related message classes.
  • replicated_value.py supports three functions: track and create PaxosInstance, link Messenger and PaxosInstance, and save and restore state.
  • resolution_strategy.py ensures that a paxos instance will achieve resolution.
  • sync_strategy.py allows a server to send messages to a random peer periodically for synchronization.
  • master_strategy.py supports master leases.

Reference

• multi-paxos: Example multi-paxos application for those learning Paxos & multi-paxos.