""" Starting in CherryPy 3.1, cherrypy.server is implemented as an :ref:`Engine Plugin`. It's an instance of :class:`cherrypy._cpserver.Server`, which is a subclass of :class:`cherrypy.process.servers.ServerAdapter`. The ``ServerAdapter`` class is designed to control other servers, as well. Multiple servers/ports ====================== If you need to start more than one HTTP server (to serve on multiple ports, or protocols, etc.), you can manually register each one and then start them all with engine.start:: s1 = ServerAdapter( cherrypy.engine, MyWSGIServer(host='0.0.0.0', port=80) ) s2 = ServerAdapter( cherrypy.engine, another.HTTPServer(host='127.0.0.1', SSL=True) ) s1.subscribe() s2.subscribe() cherrypy.engine.start() .. index:: SCGI FastCGI/SCGI ============ There are also Flup\ **F**\ CGIServer and Flup\ **S**\ CGIServer classes in :mod:`cherrypy.process.servers`. To start an fcgi server, for example, wrap an instance of it in a ServerAdapter:: addr = ('0.0.0.0', 4000) f = servers.FlupFCGIServer(application=cherrypy.tree, bindAddress=addr) s = servers.ServerAdapter(cherrypy.engine, httpserver=f, bind_addr=addr) s.subscribe() The :doc:`cherryd` startup script will do the above for you via its `-f` flag. Note that you need to download and install `flup `_ yourself, whether you use ``cherryd`` or not. .. _fastcgi: .. index:: FastCGI FastCGI ------- A very simple setup lets your cherry run with FastCGI. You just need the flup library, plus a running Apache server (with ``mod_fastcgi``) or lighttpd server. CherryPy code ^^^^^^^^^^^^^ hello.py:: #!/usr/bin/python import cherrypy class HelloWorld: '''Sample request handler class.''' @cherrypy.expose def index(self): return "Hello world!" cherrypy.tree.mount(HelloWorld()) # CherryPy autoreload must be disabled for the flup server to work cherrypy.config.update({'engine.autoreload.on':False}) Then run :doc:`/deployguide/cherryd` with the '-f' arg:: cherryd -c -d -f -i hello.py Apache ^^^^^^ At the top level in httpd.conf:: FastCgiIpcDir /tmp FastCgiServer /path/to/cherry.fcgi -idle-timeout 120 -processes 4 And inside the relevant VirtualHost section:: # FastCGI config AddHandler fastcgi-script .fcgi ScriptAliasMatch (.*$) /path/to/cherry.fcgi$1 Lighttpd ^^^^^^^^ For `Lighttpd `_ you can follow these instructions. Within ``lighttpd.conf`` make sure ``mod_fastcgi`` is active within ``server.modules``. Then, within your ``$HTTP["host"]`` directive, configure your fastcgi script like the following:: $HTTP["url"] =~ "" { fastcgi.server = ( "/" => ( "script.fcgi" => ( "bin-path" => "/path/to/your/script.fcgi", "socket" => "/tmp/script.sock", "check-local" => "disable", "disable-time" => 1, "min-procs" => 1, "max-procs" => 1, # adjust as needed ), ), ) } # end of $HTTP["url"] =~ "^/" Please see `Lighttpd FastCGI Docs `_ for an explanation of the possible configuration options. """ import os import sys import time import warnings class ServerAdapter(object): """Adapter for an HTTP server. If you need to start more than one HTTP server (to serve on multiple ports, or protocols, etc.), you can manually register each one and then start them all with bus.start:: s1 = ServerAdapter(bus, MyWSGIServer(host='0.0.0.0', port=80)) s2 = ServerAdapter(bus, another.HTTPServer(host='127.0.0.1', SSL=True)) s1.subscribe() s2.subscribe() bus.start() """ def __init__(self, bus, httpserver=None, bind_addr=None): self.bus = bus self.httpserver = httpserver self.bind_addr = bind_addr self.interrupt = None self.running = False def subscribe(self): self.bus.subscribe('start', self.start) self.bus.subscribe('stop', self.stop) def unsubscribe(self): self.bus.unsubscribe('start', self.start) self.bus.unsubscribe('stop', self.stop) def start(self): """Start the HTTP server.""" if self.running: self.bus.log('Already serving on %s' % self.description) return self.interrupt = None if not self.httpserver: raise ValueError('No HTTP server has been created.') if not os.environ.get('LISTEN_PID', None): # Start the httpserver in a new thread. if isinstance(self.bind_addr, tuple): wait_for_free_port(*self.bind_addr) import threading t = threading.Thread(target=self._start_http_thread) t.setName('HTTPServer ' + t.getName()) t.start() self.wait() self.running = True self.bus.log('Serving on %s' % self.description) start.priority = 75 @property def description(self): """ A description about where this server is bound. """ if self.bind_addr is None: on_what = 'unknown interface (dynamic?)' elif isinstance(self.bind_addr, tuple): on_what = self._get_base() else: on_what = 'socket file: %s' % self.bind_addr return on_what def _get_base(self): if not self.httpserver: return '' host, port = self.bound_addr if getattr(self.httpserver, 'ssl_adapter', None): scheme = 'https' if port != 443: host += ':%s' % port else: scheme = 'http' if port != 80: host += ':%s' % port return '%s://%s' % (scheme, host) def _start_http_thread(self): """HTTP servers MUST be running in new threads, so that the main thread persists to receive KeyboardInterrupt's. If an exception is raised in the httpserver's thread then it's trapped here, and the bus (and therefore our httpserver) are shut down. """ try: self.httpserver.start() except KeyboardInterrupt: self.bus.log(' hit: shutting down HTTP server') self.interrupt = sys.exc_info()[1] self.bus.exit() except SystemExit: self.bus.log('SystemExit raised: shutting down HTTP server') self.interrupt = sys.exc_info()[1] self.bus.exit() raise except: self.interrupt = sys.exc_info()[1] self.bus.log('Error in HTTP server: shutting down', traceback=True, level=40) self.bus.exit() raise def wait(self): """Wait until the HTTP server is ready to receive requests.""" while not getattr(self.httpserver, 'ready', False): if self.interrupt: raise self.interrupt time.sleep(.1) # Wait for port to be occupied if not os.environ.get('LISTEN_PID', None): # Wait for port to be occupied if not running via socket-activation # (for socket-activation the port will be managed by systemd ) if isinstance(self.bind_addr, tuple): wait_for_occupied_port(*self.bound_addr) @property def bound_addr(self): """ The bind address, or if it's an ephemeral port and the socket has been bound, return the actual port bound. """ host, port = self.bind_addr if port == 0 and self.httpserver.socket: # Bound to ephemeral port. Get the actual port allocated. port = self.httpserver.socket.getsockname()[1] return host, port def stop(self): """Stop the HTTP server.""" if self.running: # stop() MUST block until the server is *truly* stopped. self.httpserver.stop() # Wait for the socket to be truly freed. if isinstance(self.bind_addr, tuple): wait_for_free_port(*self.bound_addr) self.running = False self.bus.log('HTTP Server %s shut down' % self.httpserver) else: self.bus.log('HTTP Server %s already shut down' % self.httpserver) stop.priority = 25 def restart(self): """Restart the HTTP server.""" self.stop() self.start() class FlupCGIServer(object): """Adapter for a flup.server.cgi.WSGIServer.""" def __init__(self, *args, **kwargs): self.args = args self.kwargs = kwargs self.ready = False def start(self): """Start the CGI server.""" # We have to instantiate the server class here because its __init__ # starts a threadpool. If we do it too early, daemonize won't work. from flup.server.cgi import WSGIServer self.cgiserver = WSGIServer(*self.args, **self.kwargs) self.ready = True self.cgiserver.run() def stop(self): """Stop the HTTP server.""" self.ready = False class FlupFCGIServer(object): """Adapter for a flup.server.fcgi.WSGIServer.""" def __init__(self, *args, **kwargs): if kwargs.get('bindAddress', None) is None: import socket if not hasattr(socket, 'fromfd'): raise ValueError( 'Dynamic FCGI server not available on this platform. ' 'You must use a static or external one by providing a ' 'legal bindAddress.') self.args = args self.kwargs = kwargs self.ready = False def start(self): """Start the FCGI server.""" # We have to instantiate the server class here because its __init__ # starts a threadpool. If we do it too early, daemonize won't work. from flup.server.fcgi import WSGIServer self.fcgiserver = WSGIServer(*self.args, **self.kwargs) # TODO: report this bug upstream to flup. # If we don't set _oldSIGs on Windows, we get: # File "C:\Python24\Lib\site-packages\flup\server\threadedserver.py", # line 108, in run # self._restoreSignalHandlers() # File "C:\Python24\Lib\site-packages\flup\server\threadedserver.py", # line 156, in _restoreSignalHandlers # for signum,handler in self._oldSIGs: # AttributeError: 'WSGIServer' object has no attribute '_oldSIGs' self.fcgiserver._installSignalHandlers = lambda: None self.fcgiserver._oldSIGs = [] self.ready = True self.fcgiserver.run() def stop(self): """Stop the HTTP server.""" # Forcibly stop the fcgi server main event loop. self.fcgiserver._keepGoing = False # Force all worker threads to die off. self.fcgiserver._threadPool.maxSpare = ( self.fcgiserver._threadPool._idleCount) self.ready = False class FlupSCGIServer(object): """Adapter for a flup.server.scgi.WSGIServer.""" def __init__(self, *args, **kwargs): self.args = args self.kwargs = kwargs self.ready = False def start(self): """Start the SCGI server.""" # We have to instantiate the server class here because its __init__ # starts a threadpool. If we do it too early, daemonize won't work. from flup.server.scgi import WSGIServer self.scgiserver = WSGIServer(*self.args, **self.kwargs) # TODO: report this bug upstream to flup. # If we don't set _oldSIGs on Windows, we get: # File "C:\Python24\Lib\site-packages\flup\server\threadedserver.py", # line 108, in run # self._restoreSignalHandlers() # File "C:\Python24\Lib\site-packages\flup\server\threadedserver.py", # line 156, in _restoreSignalHandlers # for signum,handler in self._oldSIGs: # AttributeError: 'WSGIServer' object has no attribute '_oldSIGs' self.scgiserver._installSignalHandlers = lambda: None self.scgiserver._oldSIGs = [] self.ready = True self.scgiserver.run() def stop(self): """Stop the HTTP server.""" self.ready = False # Forcibly stop the scgi server main event loop. self.scgiserver._keepGoing = False # Force all worker threads to die off. self.scgiserver._threadPool.maxSpare = 0 def client_host(server_host): """Return the host on which a client can connect to the given listener.""" if server_host == '0.0.0.0': # 0.0.0.0 is INADDR_ANY, which should answer on localhost. return '127.0.0.1' if server_host in ('::', '::0', '::0.0.0.0'): # :: is IN6ADDR_ANY, which should answer on localhost. # ::0 and ::0.0.0.0 are non-canonical but common # ways to write IN6ADDR_ANY. return '::1' return server_host def check_port(host, port, timeout=1.0): """Raise an error if the given port is not free on the given host.""" if not host: raise ValueError("Host values of '' or None are not allowed.") host = client_host(host) port = int(port) import socket # AF_INET or AF_INET6 socket # Get the correct address family for our host (allows IPv6 addresses) try: info = socket.getaddrinfo(host, port, socket.AF_UNSPEC, socket.SOCK_STREAM) except socket.gaierror: if ':' in host: info = [( socket.AF_INET6, socket.SOCK_STREAM, 0, '', (host, port, 0, 0) )] else: info = [(socket.AF_INET, socket.SOCK_STREAM, 0, '', (host, port))] for res in info: af, socktype, proto, canonname, sa = res s = None try: s = socket.socket(af, socktype, proto) # See http://groups.google.com/group/cherrypy-users/ # browse_frm/thread/bbfe5eb39c904fe0 s.settimeout(timeout) s.connect((host, port)) s.close() except socket.error: if s: s.close() else: raise IOError('Port %s is in use on %s; perhaps the previous ' 'httpserver did not shut down properly.' % (repr(port), repr(host))) # Feel free to increase these defaults on slow systems: free_port_timeout = 0.1 occupied_port_timeout = 1.0 def wait_for_free_port(host, port, timeout=None): """Wait for the specified port to become free (drop requests).""" if not host: raise ValueError("Host values of '' or None are not allowed.") if timeout is None: timeout = free_port_timeout for trial in range(50): try: # we are expecting a free port, so reduce the timeout check_port(host, port, timeout=timeout) except IOError: # Give the old server thread time to free the port. time.sleep(timeout) else: return raise IOError('Port %r not free on %r' % (port, host)) def wait_for_occupied_port(host, port, timeout=None): """Wait for the specified port to become active (receive requests).""" if not host: raise ValueError("Host values of '' or None are not allowed.") if timeout is None: timeout = occupied_port_timeout for trial in range(50): try: check_port(host, port, timeout=timeout) except IOError: # port is occupied return else: time.sleep(timeout) if host == client_host(host): raise IOError('Port %r not bound on %r' % (port, host)) # On systems where a loopback interface is not available and the # server is bound to all interfaces, it's difficult to determine # whether the server is in fact occupying the port. In this case, # just issue a warning and move on. See issue #1100. msg = 'Unable to verify that the server is bound on %r' % port warnings.warn(msg)