This is an implementation for my design on a Peer-to-Peer messaging app with strong consistency.
The message log that appears for each user will always be complete and in order. On restart after faults or recovery after partitions, the user will fast-forward to the full complete log. It turns out the final design closely resembles Raft.
V - View this node's status info, such as the node's port, name, connections, stage.
M - View Message Log of chat to the knowledge of this node. Available only in Stage 3.
N - Advances stage. Can only be executed by leader.
- For each node you want to simulate, run
python3 main.pyin separate terminal windows. - It will ask for the following:
- Enter your node's port (8001-8005): Enter any value, suggest to use 8000~. Make sure each node uses a different port.
- Enter your name: Enter any name for the node
- Enter the node to connect to: Enter 0 or own port to be host, or the port number of the host.
- Once some nodes have joined, host should type
Nto advance to messaging stage. - Nodes can start typing in their messages. Due to the implementation only available in CLI, the full message log can only be viewable by typing
M.
4 Stages:
- Discovery Phase - Nodes should join at this time
- Pre-leader election phase - Confirms nodes are alive (automatic step)
- Leader election phase - Use Paxos to randomly select leader.
- Message Exchange Phase - Normal messaging app behavior. Type
Mto view messsage log. New nodes can still join at this time.