zerorpc is a flexible RPC implementation based on zeromq and messagepack. Service APIs exposed with zerorpc are called "zeroservices".
zerorpc can be used programmatically or from the command-line. It comes with a convenient script, "zerorpc", allowing to:
- expose Python modules without modifying a single line of code,
- call those modules remotely through the command line.
On most systems, its a matter of:
$ pip install zerorpc
Depending of the support from Gevent and PyZMQ on your system, you might need to install libev (for gevent) and libzmq (for pyzmq) with the development files.
Create a server with a one-liner
Let's see zerorpc in action with a simple example. In a first terminal, we will expose the Python "time" module:
$ zerorpc --server --bind tcp://*:1234 time
The bind address uses the zeromq address format. You are not limited to TCP transport: you could as well specify ipc:///tmp/time to use host-local sockets, for instance. "tcp://*:1234" is a short-hand to "tcp://0.0.0.0:1234" and means "listen on TCP port 1234, accepting connections on all IP addresses".
Call the server from the command-line
Now, in another terminal, call the exposed module:
$ zerorpc --client --connect tcp://127.0.0.1:1234 strftime %Y/%m/%d Connecting to "tcp://127.0.0.1:1234" "2011/03/07"
Since the client usecase is the most common one, "--client" is the default parameter, and you can remove it safely:
$ zerorpc --connect tcp://127.0.0.1:1234 strftime %Y/%m/%d Connecting to "tcp://127.0.0.1:1234" "2011/03/07"
Moreover, since the most common usecase is to connect (as opposed to bind) you can also omit "--connect":
$ zerorpc tcp://127.0.0.1:1234 strftime %Y/%m/%d Connecting to "tcp://127.0.0.1:1234" "2011/03/07"
See remote service documentation
You can introspect the remote service; it happens automatically if you don't specify the name of the function you want to call:
$ zerorpc tcp://127.0.0.1:1234 Connecting to "tcp://127.0.0.1:1234" tzset tzset(zone) ctime ctime(seconds) -> string clock clock() -> floating point number struct_time <undocumented> time time() -> floating point number strptime strptime(string, format) -> struct_time gmtime gmtime([seconds]) -> (tm_year, tm_mon, tm_mday, tm_hour, tm_min, mktime mktime(tuple) -> floating point number sleep sleep(seconds) asctime asctime([tuple]) -> string strftime strftime(format[, tuple]) -> string localtime localtime([seconds]) -> (tm_year,tm_mon,tm_mday,tm_hour,tm_min,
Specifying non-string arguments
Now, see what happens if we try to call a function expecting a non-string argument:
$ zerorpc tcp://127.0.0.1:1234 sleep 3 Connecting to "tcp://127.0.0.1:1234" Traceback (most recent call last): [...] TypeError: a float is required
That's because all command-line arguments are handled as strings. Don't worry, we can specify any kind of argument using JSON encoding:
$ zerorpc --json tcp://127.0.0.1:1234 sleep 3 Connecting to "tcp://127.0.0.1:1234" [wait for 3 seconds...] null
zeroworkers: reversing bind and connect
Sometimes, you don't want your client to connect to the server; you want your server to act as a kind of worker, and connect to a hub or queue which will dispatch requests. You can achieve this by swapping "--bind" and "--connect":
$ zerorpc --bind tcp://*:1234 strftime %Y/%m/%d
We now have "something" wanting to call the "strftime" function, and waiting for a worker to connect to it. Let's start the worker:
$ zerorpc --server tcp://127.0.0.1:1234 time
The worker will connect to the listening client and ask him "what should I do?"; the client will send the "strftime" function call; the worker will execute it and return the result. The first program will display the local time and exit. The worker will remain running.
Listening on multiple addresses
What if you want to run the same server on multiple addresses? Just repeat the "--bind" option:
$ zerorpc --server --bind tcp://*:1234 --bind ipc:///tmp/time time
You can then connect to it using either "zerorpc tcp://*:1234" or "zerorpc ipc:///tmp/time".
Wait, there is more! You can even mix "--bind" and "--connect". That means that your server will wait for requests on a given address, and connect as a worker on another. Likewise, you can specify "--connect" multiple times, so your worker will connect to multiple queues. If a queue is not running, it won't affect the worker (that's the magic of zeromq).
A client should probably not connect to multiple addresses!
Almost all other scenarios will work; but if you ask a client to connect to multiple addresses, and at least one of them has no server at the end, the client will ultimately block. A client can, however, bind multiple addresses, and will dispatch requests to available workers. If you want to connect to multiple remote servers for high availability purposes, you insert something like HAProxy in the middle.
Exposing a zeroservice programmatically
Of course, the command-line is simply a convenience wrapper for the zerorpc python API. Below are a few examples.
Here's how to expose an object of your choice as a zeroservice:
class Cooler(object): """ Various convenience methods to make things cooler. """ def add_man(self, sentence): """ End a sentence with ", man!" to make it sound cooler, and return the result. """ return sentence + ", man!" def add_42(self, n): """ Add 42 to an integer argument to make it cooler, and return the result. """ return n + 42 def boat(self, sentence): """ Replace a sentence with "I'm on a boat!", and return that, because it's cooler. """ return "I'm on a boat!" import zerorpc s = zerorpc.Server(Cooler()) s.bind("tcp://0.0.0.0:4242") s.run()
Let's save this code to cooler.py and run it:
$ python cooler.py
Now, in another terminal, let's try connecting to our awesome zeroservice:
$ zerorpc -j tcp://localhost:4242 add_42 1 43 $ zerorpc tcp://localhost:4242 add_man 'I own a mint-condition Volkswagen Golf' "I own a mint-condition Volkswagen Golf, man!" $ zerorpc tcp://localhost:4242 boat 'I own a mint-condition Volkswagen Golf, man!' "I'm on a boat!"
Congratulations! You have just made the World a little cooler with your first zeroservice, man!