pkrumins / upnode

transactional connection queue for dnode

upnode

Keep a dnode connection alive and re-establish state between reconnects with a transactional message queue.

examples

simple service interruption

Write a dnode server as usual but enable the upnode ping middleware with server.use(upnode.ping).

server.js:

var upnode = require('upnode');

var server = dnode(function (client, conn) {
    this.time = function (cb) { cb(new Date().toString()) };
});
server.listen(7000);

Now when you want to make a call to the server, guard your connection in the up() function. If the connection is alive the callback fires immediately. If the connection is down the callback is buffered and fires when the connection is ready again.

client.js:

var upnode = require('upnode');
var up = upnode.connect(7000);

setInterval(function () {
    up(function (remote) {
        remote.time(function (t) {
            console.log('time = ' + t);
        });
    });
}, 1000);

If we fire the client up first, then wait a few seconds to fire up the server:

$ node client.js & sleep 5; node server.js
[1] 9165
time = Fri Dec 16 2011 23:47:48 GMT-0800 (PST)
time = Fri Dec 16 2011 23:47:48 GMT-0800 (PST)
time = Fri Dec 16 2011 23:47:48 GMT-0800 (PST)
time = Fri Dec 16 2011 23:47:48 GMT-0800 (PST)
time = Fri Dec 16 2011 23:47:48 GMT-0800 (PST)
time = Fri Dec 16 2011 23:47:49 GMT-0800 (PST)
time = Fri Dec 16 2011 23:47:50 GMT-0800 (PST)
time = Fri Dec 16 2011 23:47:51 GMT-0800 (PST)
time = Fri Dec 16 2011 23:47:52 GMT-0800 (PST)

we can see that the first 5 seconds worth of requests are buffered and all come through at 23:47:48. The requests then come in one per second once the connection has been established.

If we kill the server and bring it back again while the client is running we can observe a similar discontinuity as all the pending requests come through at 23:50:20:

$ node client.js 
time = Fri Dec 16 2011 23:50:11 GMT-0800 (PST)
time = Fri Dec 16 2011 23:50:11 GMT-0800 (PST)
time = Fri Dec 16 2011 23:50:12 GMT-0800 (PST)
time = Fri Dec 16 2011 23:50:13 GMT-0800 (PST)
time = Fri Dec 16 2011 23:50:20 GMT-0800 (PST)
time = Fri Dec 16 2011 23:50:20 GMT-0800 (PST)
time = Fri Dec 16 2011 23:50:20 GMT-0800 (PST)
time = Fri Dec 16 2011 23:50:20 GMT-0800 (PST)
time = Fri Dec 16 2011 23:50:20 GMT-0800 (PST)
time = Fri Dec 16 2011 23:50:20 GMT-0800 (PST)
time = Fri Dec 16 2011 23:50:20 GMT-0800 (PST)
time = Fri Dec 16 2011 23:50:21 GMT-0800 (PST)
time = Fri Dec 16 2011 23:50:22 GMT-0800 (PST)

authenticated interruption

Oftentimes you'll want to re-establish state between reconnection attempts.

Suppose we have a simple dnode server with a beep function protected behind an auth function:

server.js:

var upnode = require('upnode');

var server = upnode(function (client, conn) {
    this.auth = function (user, pass, cb) {
        if (user === 'moo' && pass === 'hax') {
            cb(null, {
                beep : function (fn) { fn('boop at ' + new Date) }
            });
        }
        else cb('ACCESS DENIED')
    };
});
server.listen(7000);

Now instead of doing remote.auth() every time the connection drops, we can pass in a callback to upnode.connect() that will handle the re-authentication and expose the authenticated object to the up() transaction:

client.js:

var upnode = require('upnode');
var up = upnode.connect(7000, function (remote, conn) {
    remote.auth('moo', 'hax', function (err, res) {
        if (err) console.error(err)
        else conn.emit('up', res)
    });
});

setInterval(function () {
    up(function (remote) {
        remote.beep(function (s) {
            console.log(s);
        });
    });
}, 1000);

Now spin up the client.js and the server.js:

$ node client.js & sleep 2; node server.js
[1] 10892
boop at Sat Dec 17 2011 01:30:15 GMT-0800 (PST)
boop at Sat Dec 17 2011 01:30:15 GMT-0800 (PST)
boop at Sat Dec 17 2011 01:30:16 GMT-0800 (PST)
boop at Sat Dec 17 2011 01:30:17 GMT-0800 (PST)
boop at Sat Dec 17 2011 01:30:18 GMT-0800 (PST)

Kill the server a few times and observe that the client re-authenticates between reconnects.

You could do any other sort of stateful operation here besides authentication. Just emit the object you want to expose to up() through conn.emit('up', obj).

methods

var upnode = require('upnode')

var up = upnode(constructor={}).connect(...)

Establish a new dnode-style connection with the dnode function or object constructor and the connection parameters which may contain host strings, port numbers, option objects, and a connection callback in any order.

Returns a transaction function up() for the connection.

The up object emits "up" when the link is established, "down" when the link is severed, and "reconnect"` for each reconnection attempt.

If you give .connect() a callback, you must emit an 'up', remote event on the conn object with the remote object you want to make available to the subsequent up() transactions.

If you don't pass a callback to .connect() this default callback is used:

function (remote, conn) {
    conn.emit('up', remote);
}

The callback must emit an 'up' event so that state can be rebuilt between connection interruptions. A great use for this behavior is authentication where certain functionality is only made available through the callback to a .auth(username, password, cb) function on the remote. For that case you could write a connection callback that looks like:

function (remote, conn) {
    remote.auth(user, pass, function (err, obj) {
        if (err) console.error(err)
        else conn.emit('up', obj)
    });
}

and your dnode sessions will be re-authenticated between reconnects. The remote object handle in up() will be the obj result provided by the auth() callback.

Besides being passed directly to dnode's .connect(...), these additional option-object arguments are respected:

• ping - Interval in milliseconds to send pings to the remote server. Default 10000. Set to 0 to disable pings.
• timeout - Time in milliseconds to wait for a ping response before triggering a reconnect. Default 5000.
• reconnect - Time in milliseconds to wait between reconnection attempts. Default 1000.

var up = upnode.connect(...)

Shortcut for upnode({}).connect(...) like how dnode.connect(...) is a shortcut for dnode({}).connect(...).

up(timeout=0, cb)

Create a new transaction from the callback cb.

If the connection is ready, cb(remote, conn) will fire immediately. Otherwise cb will be queued until the connection is available again.

If timeout is specified, fire cb() after timeout milliseconds with no arguments. Here's an example of using timeouts:

up(5000, function (remote) {
    if (!remote) console.error('resource timed out')
    else remote.beep()
})

up.close()

Close the connection and don't attempt to reconnect.

upnode(cons).listen(...)

Create and return a new dnode server with ping middleware loaded that pings the connected clients to ensure connectivity to all connected nodes.

upnode.ping()

Middleware that sets this.ping to function (cb) { cb() }.

In your dnode server, do server.use(upnode.ping).

install

With npm do:

npm install upnode

license

MIT/X11