A fully featured Memcached client build on top of Node.js. Build with scaling in mind so it will support Memcached clusters and consistent hashing.
Pull request Compare This branch is 282 commits behind 3rd-Eden:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.



memcached is a fully featured Memcached client for Node.js. memcached is build with scaling, high availability and exceptional performance in mind. We use consistent hashing to store the data across different nodes. Consistent hashing is a scheme that provides a hash table functionality in a way that adding or removing a server node does not significantly change the mapping of the keys to server nodes. The algorithm that is used for consistent hashing is the same as libketama.

There are different ways to handle errors for example, when a server becomes unavailable you can configure the client to see all requests to that server as cache misses until it goes up again. It's also possible to automatically remove the affected server from the consistent hashing algorithm or provide memcached with a failover server that can take the place of the unresponsive server.

When these issues occur the memcached client will emit different events where you can subscribe to containing detailed information about the issues.

The client is configurable on different levels. There's a global configuration that you update so all you Memcached clusters will use the same failure configuration for example, but it's also possible to overwrite these changes per memcached instance.

Setting up the client

The constructor of the memcached client take 2 different arguments server locations and options. Syntax:

var Memcached = require('memcached');
var memcached = new Memcached(Server locations, options);

Server locations

The server locations is designed to work with different formats. These formats are all internally parsed to the correct format so our consistent hashing scheme can work with it. You can either use:

  1. String, this only works if you have are running a single server instance of Memcached. It's as easy a suppling a string in the following format: hostname:port. For example This would tell the client to connect to host on port number 11212.

  2. Array, if you are running a single server you would only have to supply one item in the array. The array format is particularly useful if you are running a cluster of Memcached servers. This will allow you to spread the keys and load between the different servers. Giving you higher availability for when one of your Memcached servers goes down.

  3. Object, when you are running a cluster of Memcached servers it could happen to not all server can allocate the same amount of memory. You might have a Memcached server with 128mb, 512, 128mb. If you would the array structure all servers would have the same weight in the consistent hashing scheme. Spreading the keys 33/33/33 over the servers. But as server 2 has more memory available you might want to give it more weight so more keys get stored on that server. When you are using a object, the key should represent the server location syntax and the value the weight of the server. By default all servers have a weight of 1. { '': 1, '': 2, '': 1 } would generate a 25/50/25 distribution of the keys.

If you would implement one of the above formats, your constructor would something like this:

var memcached = new Memcached({ '': 1, '': 2, '': 1 });
var memcached = new Memcached([ '', '', '' ]);
var memcached = new Memcached('');


There 2 kinds of options that can be configured. A global configuration that will be inherited by all Memcached servers instances and a client specific configuration that can be used to overwrite the globals. The options should be formatted in an JavaScript object. They both use the same object structure:

  • maxKeySize: 250, the max size of they key allowed by the Memcached server.
  • maxExpiration: 2592000, the max expiration of keys by the Memcached server in milliseconds.
  • maxValue: 1048576, the max size of a value that is allowed by the Memcached server.
  • poolSize: 10, the maximum connections we can allocate in our connection pool.
  • algorithm: crc32, the hashing algorithm that should be used to generate the hashRing values.
  • reconnect: 18000000, when the server is marked as dead we will attempt to reconnect every x milliseconds.
  • timeout: 5000, after x ms the server should send a timeout if we can't connect. This will also be used close the connection if we are idle.
  • retries: 5, amount of tries before we mark the server as dead.
  • retry: 30000, timeout between each retry in x milliseconds.
  • remove: false, when the server is marked as dead you can remove it from the pool so all other will receive the keys instead.
  • failOverServers: undefined, the ability use these servers as failover when the dead server get's removed from the consistent hashing scheme. This must be an array of servers confirm the server_locations specification.
  • keyCompression: true, compress keys using md5 if they exceed the maxKeySize option.

Example usage:

var memcached = new Memcached('localhost:11212', {retries:10,retry:10000,remove:true,failOverServers:['']});

If you wish to configure the options globally:

var Memcached = require('memcached');
// all global configurations should be applied to the .config object of the Client.
Memcached.config.poolSize = 25;


Public methods

Private methods

The following methods are intended for private usage:


Fetches or generates a connection for the given server. The supplied callback function will receive a reference to the connection as argument. If there are issues with the server connection, we are going to respond with cache-miss pattern.


server: String, The server that needs a connection, the format must be confirm the server_locations specification.

callback: Function, The callback function that receives the net.Stream connection. It will be called with 2 arguments error and connection.


memcached.connect( '', function( err, conn ){
	if( err ) throw new Error( err );
	console.log( conn.server );


A small wrapper function that makes it easier to query multiple Memcached servers. It will return the location for each key or the complete list of servers.


keys: Array (optional), They keys that needs to be converted to a server.

callback: Function, The callback function for the data, it will be called for each key. It will be called with 4 arguments:

  1. server: String, The server location.
  2. key: String, The key associated with the server, if you didn't specify keys, this variable will be undefined.
  3. index: Number, The current index of the loop
  4. total: Number, The total amount server retrieved.


memcached.multi( false, function( server, key, index, totals ){
	if( err ) throw new Error( err );
	this.connect( server, function( err, conn ){
		console.log( "connection ready" )


This is the core functionality of the memcached client. All public API's are routed through this function. It takes care of the argument validations Server retrieval ( If the server argument isn't specified ). After all data ready a connection is asked for the private connect method and the command is written to the Memcached server.


query: Object, The metaData object, see the Callbacks section for the specification.

server: String, The server the to connect. This is only needed when the metaData object doesn't contain a key property to retrieve the server from.


	key: 'key', callback: function(){ console.dir( arguments ); },

	// validate the arguments
	validate: [[ 'key', String ], [ 'callback', Function ]],

	// used for the query
	type: 'delete',
	command: 'delete key'


A internal function for logging issues with connections. As there can be various of ways that an error occurs we need solid issue manager to handle all these cases. For example server could crash or the Memcached server could respond with SERVER ERROR <broken>.


error: String, The actual error message.

Stream: net.Stream, A reference to the connection stream where the error occurred on.

callback: Function (optional), The callback function of a potential request, it will be marked as cache miss if it was provided


memcached.connectionIssue( "Server down", connectionReference );


Each method requires a callback function. Once this function get executed there will be 2 variables applied:

  • error: A error response if something went wrong while retrieving data from the Memcached server. Depending on the type of request this will either be an string or an Array with multiple errors.
  • response: The actual result from the Memcached server. If the response is false or undefined than a cache miss occurred. Cache misses will also occur when there is an error. So you might want to check on errors first.

When we have a successful response, the context of the callback function will shift to a metaData object. The metaData object contains all information that we used to generate the request for the Memcached server. The metaData object contains the following properties:

  • start: Date in milliseconds when the request was received
  • execution: Total execution time for the request, including response parsing.
  • callback: Reference to the callback function
  • type: The type of Memcached command
  • command: The compiled command that was send through the sockets
  • validate: The properties of metaData object that needs type validation.

And all the arguments you have send to the method, this depends on the method you have called.


When connection issues occur we send out different notifications using the EventEmitter protocol. This can be useful for logging, notification and debugging purposes. Each event will receive details Object containing detailed information about the issues that occurred.

Details Object

The details Object contains the various of error messages that caused, the following 3 will always be present in all error events:

  • server: the server where the issue occurred on
  • tokens: a array of the parsed server string in [port, hostname] format.
  • messages: a array containing all error messages that this server received. As messages are added to the array using .push(), the first issue will at the beginning and the latest error at the end of the array.

The following properties depend on the type of event that is send. If we are still in our retry phase the details will also contain:

  • retries: the amount of retries left before we mark the server as dead.
  • totalRetries: the total amount of retries we did on this server, as when the server has been reconnected after it's dead the retries will be rest to defaults and messages will be removed.

If the server is dead these details will be added:

  • totalReconnectsAttempted: the total reconnects we have attempted. This is the success and failure combined.
  • totalReconnectsSuccess: the total successful reconnects we have made.
  • totalReconnectsFailed: the total failed reconnects we have made.
  • totalDownTime: the total down time that was generated. Formula: ( totalReconnectsFailed * reconnect_timeout ) + ( totalRetries * retry_timeout).


There are 5 different events that the memcached client emits when connection issues occur.

  • issue: a issue occurred on one a server, we are going to attempt a retry next.
  • failure: a server has been marked as failure or dead.
  • reconnecting: we are going to attempt to reconnect the to the failed server.
  • reconnected: successfully reconnected to the memcached server.
  • remove: removing the server from our consistent hashing.

Example implementations:

var memcached = new Memcached([ '', '' ]);
memcached.on('failure', function( details ){ sys.error( "Server " + details.server + "went down due to: " + details.messages.join( '' ) ) });
memcached.on('reconnecting', function( details ){ sys.debug( "Total downtime caused by server " + details.server + " :" + details.totalDownTime + "ms")});