Skip to content

Latest commit

 

History

History
83 lines (60 loc) · 2.63 KB

README.md

File metadata and controls

83 lines (60 loc) · 2.63 KB

Nsync

No, this is not a boys-band, this is a program that when ran on multiple machines, syncs files inside specified directories.

Design

Nsync is built with Python (for demo purposes) on leader-follower model with automatic leader election using an implementation of Raft in Python called PySyncObj to maintain a shared state of the cluster about modified files list.

Whenever a file is modified or deleted on a node, it is announced on the cluster state along with the list of nodes it is already synced to, originally starting with the source node, e.g.:

{
    ...

    "filename": {
        "event": "modified",
        "synced_to": ["node1"]
    }

    ...
}

A worker on each node monitors the cluster state, and whenever a file event is detected that has not been synced to the current node, the node downloads (or deletes) the file over HTTP (because of task time limit) from one of the other nodes the file is already synced to. After this is done, the current node is added to the synced_to list on the cluster state.

The same worker on a leader node cleans up the files on the cluster state that have been synced to all the nodes in the cluster.

This design allows any node to go down at any point, and then come back online and catch up with the state. Initial sync of files is not implemented due to time limit, the nodes are assumed to start on directories in identical state (or empty).

Simultaneous writes are handled through a distributed lock implemented in the same PySyncObj library. Whichever node happens to acquire the lock first, gets to announce the update to the file. The competing node picks up the event of the other node and overwrites the updated file.

Usage

Please follow the steps below to use Nsync:

  • First clone the repository:

    $ git clone https://github.com/bagrat/nsync.git

  • Install dependencies:

    $ cd nsync
$ pipenv install --dev

  • Optional: run tests

    $ pipenv run py.test -v

  • Run multiple instances (check out help with --help):

    $ pipenv run python nsync <YOUR_PATH_1> --file-server-port 8001 --cluster-port 10001 --cluster localhost:10002:8002,localhost:10003:8003
$ pipenv run python nsync <YOUR_PATH_2> --file-server-port 8002 --cluster-port 10002 --cluster localhost:10001:8001,localhost:10003:8003
$ pipenv run python nsync <YOUR_PATH_3> --file-server-port 8003 --cluster-port 10003 --cluster localhost:10001:8001,localhost:10002:8002

How to make it fast?

First and foremost, rewrite it in a language faster than Python. Additionally the file sharing could be done on the TCP level, rather than HTTP.