tabby

Tabby is an implementation of the Raft consensus algorithm.

Features

• REPL based development
• Leadership Election
• Log Replication

Database

The database is just a key value store with support for the following operations:

• set key value
• get key
• compare-and-swap key new old

Client Libraries

There is only one client at this time and it's written in clojure in this repo in the client namespace.

Building

script/build

Running

There are a few ways to runtabby

1. Kubernetes: There are some example yamls in the 'k8s' directory.
2. Command line:

For each peer machine run tabby like this: java -jar $TABBY_JAR --id $ME --peers $PEER_STRING --data-dir $DATA_DIR

• $ME: The ID of peer in question
• $PEER_STRING: A comma separated list of peers in the format of := e.g. "tabby-0:7659=0,tabby-1:7659" NOTE: tabby will prune itself out of the peer list
• $DATA_DIR: Where tabby should store it's replicated log.

Ideally run this with a daemonizer, like systemd or whatever.

TODO

• Fix inconsistent jepsen results
• HTTP client API
• HTTP health checks
• Prometheus Metrics

Jespsen testing

1. Clone the jepsen repo
2. Copy the ./jepsen directory into the jepsen repo
3. Turn up jepsen, ( I use the docker one )
4. Enter the control node
5. Go into the 'tabby' subdirectory
6. lein run test --concurrency 10

License

Copyright © 2015 Ian Quick

Distributed under the Eclipse Public License either version 1.0 or any later version.