level-master

master-slave replication with levelup.

Master/Slave replication is when you have writes to one db, and reads from another. You can have either many slaves, and a single master, or many masters, and a single slave.

Either way, master-slave replication depends on centralization, but is very simple, if it fits your usecase.

Example

Replicate from a central server to clients.

//master.js
var levelup  = require('levelup')
var SubLevel = require('level-sublevel')
var net      = require('net')
var Master   = require('level-master')
 
//setup the database.
var db = SubLevel(levelup('/tmp/example-master'))
 
//install Master plugin!
var master = Master(db, 'master')
 
//create a server, and stream data to who ever connects.
net.createServer(function (stream) {
  stream.pipe(master.createStream({tail: true})).pipe(stream)
}).listen(9999, function () {
  console.log('master db listening on 9999')
})

Then, the code for the client!

//slave.js
var levelup  = require('levelup')
var SubLevel = require('level-sublevel')
var net      = require('net')
var Master   = require('level-master')
 
var db = SubLevel(levelup('/tmp/example-slave'))
var slave = Master.Slave(db, 'slave')
 
var stream = net.connect(9999)
 
stream.pipe(slave.createStream()).pipe(stream)

Wow, that was simple.

did someone say "webscale"?

master in the middle

writes go to the master, and are then copied to many slaves. requests are load balanced across the slaves...

slave in the middle

A large amount of data is written to many masters. Each master aggregates the data (probably with a module like map-reduce), and then the aggregation is replicated into the central slave.

The central slave the applies the same aggregation again, giving you global data.

(it's important here that the data from each node does not collide. keys from each node need a different prefix or to be stored in a separate sublevel)

License

MIT