locked

distributed consensus on a single value using etcd

install

$ npm install locked

usage

First create a locker that is pointing at some nodes of your etcd cluster

var locked = require('locked')


// create a locker pointing with an etcd connection string
var locker = locked('127.0.0.1:4001,127.0.0.1:4002')

// create a node - this points to a single key and represents its value locked across the cluster
var node = locker({
	id:'node1',     // this is auto-completed if left blank
	path:'/master', // the etcd path we use for the lock
	value:10 ,      // the value for this node
	ttl:10          // we will check the lock every (ttl/2) seconds
})

// this is run when the values changes regardless of what node is elected
node.on('change', function(value, nodeid){
	console.log(value + ' was written by ' + nodeid)
})

// this is run when this node has been selected
node.on('select', function(){
	console.log('this server has been elected as the master!')
})

// this is run when this node is no longer the leader
node.on('deselect', function(){
	console.log('this server is no longer the master!')
})

// this starts the lock for this node
node.start()

you can also pass an existing etcdjs object:

var etcdjs = require('etcdjs')
var locker = locked(etcdjs('127.0.0.1:4001,127.0.0.1:4002'))

api

var locker = locked(etcdhosts)

Create a new locker pointing to some of the etcd servers in your cluster - etcdhosts can be an array or a single address string.

var node = locker(path, opts)

Create a node that represents a single key - other machines will have created nodes on the same path - they are all essentially competing.

Opts has the following keys:

• id - specify the id of this node manually, this is auto-generated if left blank
• value - the value this node will write to the registry
• ttl - the amount of seconds that each value is valid - the value is automatically refreshed every ttl/2 seconds

node.start()

Start the lock process for this node

node.stop()

Stop the lock process for this node

node.id()

Get the currently active id across the lock

node.value()

Get the currently active value across the lock

node.localid()

The id for this specific node

node.localvalue()

The value for this specific node

node.localdata()

The data string for this specific node

This is id + ':::' + value

node.isSelected()

return boolean indicating if the node is currently the elected leader

events

node.on('change', function(value, leaderid){})

This event is triggered when the lock value has changed regardless of which node was elected.

The nodeid is of the elected machine is passed as the second argument.

node.on('select', function(value, leaderid){})

This event is triggered when the node has been elected and it's value distributed to the cluster.

You can run logic in this function that should only be running on one server at a time.

node.on('deselect', function(){})

This event is triggered when the node was the leader and now is not the leader (either because another node took leadership or because we were unable to refresh to leader itself)

node.on('refresh', function(value, leaderid){})

Triggered when the currently selected node has refreshed its value

node.on('ping', function(value, leaderid){})

Triggered when the leader has checked in and confirmed its leadership

node.on('error', function(error){})

If there is an error connecting to the etcd instance or with the HTTP request for each leadership poll, this event will be called with the error.

Upon such an error - the lock will cancel its polling for leadership and will leave you to clean up / restart the lock.

license

MIT