Shard and enable pre-splitting (on hashed index) for collection(s) in a MongoDB Database.

preshard-database.js

/*
Shard collection(s) before restoring or mass importing data in them

Script has to run against a mongos instance (not mongod), with a user having:
- "clusterAdmin" privilege
- "readWrite" privilege on database to shard

For each collection to shard, you will have to estimate the "numInitialChunks" value depending on :
- The volume of data to import/insert
- The configured chunks size for your instance

Author : Henri-Maxime Ducoulombier ( [email protected] )
*/

var dbName = 'myDB', // Name of the database to work on
myAdmin = db.getSiblingDB("admin"), // Admin DB pointer to run admin commands
dropFirst = false, // Set to true to drop collection(s) first. Default is false for safety reasons
enableSharding = false, // If set to true, will enable sharding on database prior to sharding collection (required for not yet sharded databases)
disableBalancer = true, // If set to true, will disable the balancer before sharding (can take some time if balancing is in progress)
enableBalancer = false, // If set to true, will re-enable the balancer after the operation
shardCollections = [ // List of collections to shard, with settings per collection
	{"collection": "collection1", "numInitialChunks": 12, "key": {"myShardingKey": "hashed"}, "name": "shard_idx"},
	{"collection": "collection2", "numInitialChunks": 18, "key": {"myShardingKey": "hashed"}, "name": "shard_idx"},
	{"collection": "collection3", "numInitialChunks": 24, "key": {"myShardingKey": "hashed"}, "name": "shard_idx"},
];

if (disableBalancer){
	// Disable balancer first (you should always do that when backuping/restoring data)
	sh.setBalancerState(false);
}

if (enableSharding){
	// Enable sharding on DB (only if sharding is not yet enabled on database)
	db.getSiblingDB("admin").runCommand({"enableSharding": dbName});
}

// Browse collection(s) to shard
shardCollections.forEach(function(d){
	var collName = dbName + '.' + d.collection; // Complete name space of current collection to shard (DB + collection name)

	if (dropFirst){
		// Drop collection first (including indices)
		// Warning : This command obtains a write lock on the affected database and will block other operations until it has completed.
		db.getSiblingDB(dbName).getCollection(d.collection).drop();
	}
	
	// Create sharding index (required to give the index a name)
	db.getSiblingDB(dbName).getCollection(d.collection).createIndex(d.key, {"name": d.name});
	// Shard collection
	db.getSiblingDB("admin").runCommand({"shardCollection": collName, "key": d.key, "numInitialChunks": d.numInitialChunks});
});

// Re-enable balancer only if it was disabled and config says it has to be re-enabled
if (disableBalancer && enableBalancer){
	sh.setBalancerState(true);
}

// Note : after the operation is done, you can check sharding distribution with :
// db.getSiblingDB(dbName).getCollection(PUT_COLLECTION_NAME_HERE).getShardDistribution();