cpeng365/moduleserver

server dispatches requests to specified module

https://github.com/mwalsh161/ModuleServer

Tested on python 3.7 (uses all builtin libraries)

ModuleServer

The server runs as a standalone process to receive the initial client request. The client is passed off to a different process (the "worker") that manages the module requested by the client to fulfill the rest of the client's request.

The server monitors the workers and the workers monitor the module they are assigned to. If a worker dies, the server will try to restart it. If the code in a module gets modified, the worker will reload the module when it isn't busy with a client.

The server reads a config file to know which modules to load, and will monitor that config file for changes during runtime when it isn't busy. It will take care of loading/unloading modules based on the updated config.

Diagram showing the general construction of the server:

Setup

Server Setup

You can create a server.py file as follows:

import logging, os, sys
import ModuleServer.server as server

BASE_PATH = os.path.dirname(os.path.abspath(__file__))

LOGLEVEL = logging.INFO
CONFIG_PATH = os.path.join(BASE_PATH,'server.config')
SERVER_IP = 'localhost' # 0.0.0.0 to bind to all interfaces
SERVER_PORT = 36577
LOGFILE = os.path.join(BASE_PATH,'logs','server.log')

if __name__ == '__main__':
    server.main('HW Server',CONFIG_PATH,SERVER_IP,SERVER_PORT,LOGLEVEL,LOGFILE)

This file implies that there is a server.config file in the same directory as well as a folder called logs.

Config file

This just needs to be a JSON file that informs the server's workers how to load and dispatch requests to your module. Entries that have an underscore as the first character of the module name are ignored. The format is a dictionary with keys being the module names and values being a list: ["module import path", "class name or function entry point", "dispatch method in class or null if calling directly by client request"]

{
    "moduleA": [
        "mymodules.moduleA",
        "moduleA",
        "foo"
    ],
    "_moduleB": [
        "mymodules.moduleB",
        "bar",
        null
    ]
}

Directory structure for Example

myproject/
    server.py
    server.config
    mymodules/
        __init__.py
        moduleA.py
        moduleB.py
    logs/
        server.log
        server.log.1
        ...
        server.log.5

Modules

Modules don't have a particular required recipe. Here is a possible implementation to complete the example (based on the config file). moduleA.py:

import os, logging, time
logger = logging.getLogger(__name__)

class moduleA:
    def __init__(self):
        logger.debug('Here we go')
        time.sleep(3)
        logger.debug('Loaded')

    def foo(self,client_ip,fn_name,*args):
        logger.debug('Calling '+fn_name+str(args))
        return 'You successfully called the dispatching method!'

moduleB.py:

import os, logging, time
logger = logging.getLogger(__name__)

class bar:
    def fun1(self,*args):
        logger.debug('Called fun1 directly! Args: '+str(args))
        return 'No dispatching method necessary.'

Note, workers setup root logger properly, so modules can inherit and log easily at any level:

import logging
logger = logging.getLogger(__name__)
logger.info('testing123...')

Clients

Client connects with server.py on host machine to communicate with modules. Here is an example of creating a client. Also worth noting that client.py uses numpy-style docstrings.

from client import client
myclient = client(host = 'localhost')

The client.ping() method can be issued to ping the server to get the client's IP address and binding port.

print(myclient.ping())

The client.com(module, funcname='_help', *args) method can be called to communicate with a module and request it to perform a function. Here is an example of requesting "moduleA" to call "foo" with arguments ['ay', 1, False, None].

resp = myclient.com('moduleA', 'foo', 'ay', 1, False, None)
print(resp) #-> You successfully called the dispatching method!

The client.help() method can be called to get server help text.

The client.get_modules(prefix='') method will return a list of module names that are loaded. If you specify .* (e.g. myclient.get_modules('msquared')), only the modules that begin with * will be returned.

The client.reload(module) method can be issued to force a reload of the module specified.

Logging

It is also worth noting, that you can configure how this module performs logging if you wish. To mess around with this module with basic logging enabled, you should run it directly: python -i client.py. If you import it, logging is not configured and left to the user to do so.

API/Protocol

See server.py for more details. The general flow for the protocol is below: