A simple way to shard MongoDB on the app layer.
With this approach, we have a set of replica sets. Then we can select a shard for our data based on free disk space on the DB.
For this, you need to run identical clusters (with the same size of disk space and the same storage engine)
Install it first:
npm i mongo-sharded-cluster
Using it
var MongoCluster = require('mongo-sharded-cluster');
var MongoClient = require('mongodb').MongoClient;
var async = require('async');
var myCluster = new MongoCluster();
var replSetUrls = [
{name: "one", url: "mongodb://localhost/one"},
{name: "two", url: "mongodb://localhost/two"},
];
async.each(replSetUrls, function(replSetInfo, done) {
MongoClient.connect(replSetInfo.url, function(err, db) {
if(db) {
myCluster.add(replSetInfo.name, db);
}
done(err);
});
}, function(err) {
if(err) {
throw err;
}
myCluster.startDbSizeLookup(function() {
// Pick a shard
var shardName = myCluster.pickShard();
// Then you can save this shardName in where you need.
// Get the connection for the shard
var conn = myCluster.getConnection(shardName);
});
});
You can use following helper method to initialize a cluster without hassle.
First make sure you have exports following env variables.
export MONGO_SHARD_URL_one=mongodb://localhost/one
export MONGO_SHARD_URL_two=mongodb://localhost/two
Then you can get initialize a cluster like this:
var MongoCluster = require('mongo-sharded-cluster');
MongoCluster.initFromEnv(function(err, cluster) {
if(err) {
throw err;
}
// do anything with the cluster
});