skiff
v1.10.0
Published
Raft for Node.js
Downloads
70
Maintainers
Readme
Skiff
Raft Consensus Algorithm implementation for Node.js.
- Persists to LevelDB (or any database exposing a LevelDown interface).
- Exposes the cluster as a Levelup or Leveldown-compatible interface, with which you can extend using the Levelup plugins.
- Encodes messages using Msgpack
Installation
$ npm install skiff --save
Usage
const Skiff = require('skiff')
const options = {
db: require('memdown'), // in memory database
peers: [ // peer addresses
'/ip4/127.0.0.1/tcp/9491',
'/ip4/127.0.0.1/tcp/9492'
]
}
const skiff = Skiff('/ip4/127.0.0.1/tcp/9490', options)
// expose the cluster as a Levelup-compatible database
const db = skiff.levelup()
skiff.start(err => {
if (err) {
console.error('Error starting skiff node: ', err.message)
} else {
console.log('Skiff node started')
db.put('key', 'value', (err) => {
// ...
})
}
})
API
Skiff (address, options)
Returns a new skiff node.
Arguments:
address
(string, mandatory): an address in the multiaddr format (example:"/ip/127.0.0.1/tcp/5398"
).options
(object):network
(object): if you want to share the network with other skiff nodes on the same process, create a network usingSkiff.createNetwork(options)
(see below)server
(object):port
(integer): TCP port. Defaults to the port inaddress
host
(string): host name to bind the server to. Defaults to the host name in theaddress
- rpcTimeoutMS (integer, defaults to
2000
): Timeout for RPC calls. - peers (array of strings, defaults to
[]
): The addresses of the peers (also in the multiaddr format). If the database you're using is persisted to disk (which is the default), these peers will be overrridden by whatever is loaded from the latest snapshot once the node starts. levelup
(object): options to the internal Levelup database. Defaults to:
{ keyEncoding: 'utf8', valueEncoding: 'json' }
(You can use this to create a in-memory database using Memdown)
Advanced options
appendEntriesIntervalMS
(integer, defaults to100
): The interval (ms) with which a leader sendsAppendEntries
messages to the followers (ping).electionTimeoutMinMS
(integer, defaults to300
): The minimum election timeout (ms) for a node. It's the minimum time a node has to wait until noAppendEntries
message triggers an election.electionTimeoutMaxMS
(integer, defaults to600
): The maximum election timeout (ms) for a node. It's the maximum time a node has to wait until noAppendEntries
message triggers an election.installSnapshotChunkSize
(integer, defaults to10
): The maximum number of database records on eachInstallSnapshot
message.batchEntriesLimit
(integer, defaults to10
): The maximum number of log entries in aAppendEntries
message.clientRetryRPCTimeout
(integer, defaults to 200): The number of miliseconds the internal client has to wait until retryingclientMaxRetries
(integer, defaults to 10): The maximum number of times the client is allowed to retry the remote call.
skiff.start (callback)
Starts the node, initializing. Calls back with no argument when started, or with error in the first argument.
skiff.stop (callback)
Stops the node, shutting down server, disconnects from all peers and stops activity. Calls back once all this is done, or when an error is encountered, with an error in the first argument.
skiff.levelup ()
Returns a new Levelup-compatible object for you to interact with the cluster.
Node must be a leader in order to accept commands.
skiff.leveldown ()
Returns a new Leveldown-compatible object for you to interact with the cluster.
Node must be a leader in order to accept commands.
skiff.join (peerAddress, callback)
Adds a peer to the cluster. Calls back once the cluster reaches consensus, or with an error if no consensus can be reached.
Node must be leader for this to succeed.
skiff.leave (peerAddress, callback)
Removes a peer from the cluster. Calls back once the cluster reaches consensus, or with an error if no consensus can be reached.
Node must be leader for this to succeed.
skiff.stats ()
Returns some interesting stats for this node.
skiff.peers (callback)
Invokes the error-first callback function with the cluster peers and some interesting stats from each.
skiff.term ()
Returns the current term (integer).
skiff.weaken (durationMS)
Weakens the node for the duration. During this period, the node transitions to a special weakened
state, in which the node does not react to election timeouts. This period ends once it learns a new leader or the period runs out.
skiff.readConsensus(callback)
Asks for read consensus from the cluster. Calls back when there is an error (with the error as the first argument) or succeeded.
Events
A skiff instance emits the following events:
started
: once the node is started (network server is up and persisted state is loaded)warning (err)
: if a non-fatal error was encounteredconnect (peer)
: once a leader node is connected to a peerdisconnect (peer)
: once a leader node is disconnected from a peernew state (state)
: once a node changes state (possible states arefollower
,candidate
andleader
)leader
: once the node becomes the cluster leaderjoined (peerAddress)
: when a peer joined the clusterleft (peerAddress)
: whan a peer left the clusterrpc latency (ms)
: the latency for an RPC call, in milisenconds
Skiff.createNetwork (options)
Creates a network you can share amongst several Skiff nodes in the same process.
Options:
active
(object):innactivityTimeout
(integer, miliseconds, defaults to5000
): The amount of time to wait before a client connection is closed because of innactivity.
passive
(object):server
(object):port
(integer, defaults to9163
): the port the server should listen onhost
(string, defaults to"0.0.0.0"
): the interface address the server should listen toexclusive
(boolean, defaults totrue
): if true, the server is not shareable with other processes (seeServer#listen()
on Node.js docs).
Sponsors
Development of Skiff is sponsored by YLD.
License
Copyright
Copyright (c) 2016 Pedro Teixeira