Make pid handling more stable, also internally

evennia/server/amp_client.py

@@ -4,6 +4,7 @@ Portal. This module sets up the Client-side communication.
 
 """
 
+import os
 from evennia.server.portal import amp
 from twisted.internet import protocol
 from evennia.utils import logger
@@ -105,7 +106,8 @@ class AMPServerClientProtocol(amp.AMPMultiConnectionProtocol):
         super(AMPServerClientProtocol, self).connectionMade()
         # first thing we do is to request the Portal to sync all sessions
         # back with the Server side. We also need the startup mode (reload, reset, shutdown)
-        self.send_AdminServer2Portal(amp.DUMMYSESSION, operation=amp.PSYNC, info_dict=info_dict)
+        self.send_AdminServer2Portal(
+                amp.DUMMYSESSION, operation=amp.PSYNC, spid=os.getpid(), info_dict=info_dict)
 
     def data_to_portal(self, command, sessid, **kwargs):
         """

evennia/server/evennia_launcher.py

@@ -679,16 +679,14 @@ def query_status(callback=None):
             callback(response)
         else:
             pstatus, sstatus, ppid, spid, pinfo, sinfo = _parse_status(response)
-            # note - the server reports its pid, but this is likely to be a
+            print("Evennia Portal: {}{}\n        Server: {}{}".format(
-            # thread and can't be relied on, so we use the pid file instead
+                wmap[pstatus], " (pid {})".format(get_pid(PORTAL_PIDFILE, ppid)) if pstatus else "",
-            print("Portal:   {} (pid {})\nServer:   {} (pid {})".format(
+                wmap[sstatus], " (pid {})".format(get_pid(SERVER_PIDFILE, spid)) if sstatus else ""))
-                wmap[pstatus], get_pid(PORTAL_PIDFILE),
-                wmap[sstatus], get_pid(SERVER_PIDFILE)))
             _reactor_stop()
 
     def _errback(fail):
         pstatus, sstatus = False, False
-        print("Portal:   {}\nServer:   {}".format(wmap[pstatus], wmap[sstatus]))
+        print("Portal: {}\nServer: {}".format(wmap[pstatus], wmap[sstatus]))
         _reactor_stop()
 
     send_instruction(PSTATUS, None, _callback, _errback)
@@ -1355,22 +1353,23 @@ def getenv():
     return env
 
 
-def get_pid(pidfile):
+def get_pid(pidfile, default=None):
     """
     Get the PID (Process ID) by trying to access an PID file.
 
     Args:
         pidfile (str): The path of the pid file.
+        default (int, optional): What to return if file does not exist.
 
     Returns:
-        pid (str or None): The process id.
+        pid (str): The process id or `default`.
 
     """
     if os.path.exists(pidfile):
         with open(pidfile, 'r') as f:
             pid = f.read()
             return pid
-    return None
+    return default
 
 
 def del_pid(pidfile):
@@ -1418,7 +1417,7 @@ def kill(pidfile, component='Server', callback=None, errback=None, killsignal=SI
                     # We must catch and ignore the interrupt sent.
                     pass
             else:
-                # Linux can send the SIGINT signal directly
+                # Linux/Unix can send the SIGINT signal directly
                 # to the specified PID.
                 os.kill(int(pid), killsignal)
 

evennia/server/portal/amp_server.py

@@ -54,7 +54,6 @@ class AMPServerFactory(protocol.ServerFactory):
         self.protocol = AMPServerProtocol
         self.broadcasts = []
         self.server_connection = None
-        self.server_info_dict = None
         self.launcher_connection = None
         self.disconnect_callbacks = {}
         self.server_connect_callbacks = []
@@ -89,7 +88,7 @@ class AMPServerProtocol(amp.AMPMultiConnectionProtocol):
         super(AMPServerProtocol, self).connectionLost(reason)
         if self.factory.server_connection == self:
             self.factory.server_connection = None
-            self.factory.server_info_dict = None
+            self.factory.portal.server_info_dict = {}
         if self.factory.launcher_connection == self:
             self.factory.launcher_connection = None
 
@@ -112,7 +111,7 @@ class AMPServerProtocol(amp.AMPMultiConnectionProtocol):
         server_connected = bool(self.factory.server_connection and
                                 self.factory.server_connection.transport.connected)
         portal_info_dict = self.factory.portal.get_info_dict()
-        server_info_dict = self.factory.server_info_dict
+        server_info_dict = self.factory.portal.server_info_dict
         server_pid = self.factory.portal.server_process_id
         portal_pid = os.getpid()
         return (True, server_connected, portal_pid, server_pid, portal_info_dict, server_info_dict)
@@ -151,7 +150,11 @@ class AMPServerProtocol(amp.AMPMultiConnectionProtocol):
 
         """
         # start the Server
+        process = None
         with open(settings.SERVER_LOG_FILE, 'a') as logfile:
+            # we link stdout to a file in order to catch
+            # eventual errors happening before the Server has
+            # opened its logger.
             try:
                 if os.name == 'nt':
                     # Windows requires special care
@@ -159,24 +162,20 @@ class AMPServerProtocol(amp.AMPMultiConnectionProtocol):
                     process = Popen(server_twistd_cmd, env=getenv(), bufsize=-1,
                                     stdout=logfile, stderr=STDOUT,
                                     creationflags=create_no_window)
+
                 else:
                     process = Popen(server_twistd_cmd, env=getenv(), bufsize=-1,
                                     stdout=logfile, stderr=STDOUT)
             except Exception:
-                self.factory.portal.server_process_id = None
                 logger.log_trace()
-                logfile.flush()
-                return 0
-            # there is a short window before the server logger is up where we must
-            # catch the stdout of the Server or eventual tracebacks will be lost.
-            # with process.stdout as out:
-            #     logger.log_server(out.readlines())
 
-            # store the pid and launch argument for future reference
-            self.factory.portal.server_process_id = process.pid
             self.factory.portal.server_twistd_cmd = server_twistd_cmd
             logfile.flush()
+        if process:
+            # avoid zombie-process
+            process.wait()
             return process.pid
+        return 0
 
     def wait_for_disconnect(self, callback, *args, **kwargs):
         """
@@ -414,7 +413,8 @@ class AMPServerProtocol(amp.AMPMultiConnectionProtocol):
 
         elif operation == amp.PSYNC:  # portal sync
             # Server has (re-)connected and wants the session data from portal
-            self.factory.server_info_dict = kwargs.get("info_dict", {})
+            self.factory.portal.server_info_dict = kwargs.get("info_dict", {})
+            self.factory.portal.server_process_id = kwargs.get("spid", None)
             # this defaults to 'shutdown' or whatever value set in server_stop
             server_restart_mode = self.factory.portal.server_restart_mode
 

evennia/server/portal/portal.py

@@ -39,11 +39,6 @@ except Exception:
 PORTAL_SERVICES_PLUGIN_MODULES = [mod_import(module) for module in make_iter(settings.PORTAL_SERVICES_PLUGIN_MODULES)]
 LOCKDOWN_MODE = settings.LOCKDOWN_MODE
 
-PORTAL_PIDFILE = ""
-if os.name == 'nt':
-    # For Windows we need to handle pid files manually.
-    PORTAL_PIDFILE = os.path.join(settings.GAME_DIR, "server", 'portal.pid')
-
 # -------------------------------------------------------------
 # Evennia Portal settings
 # -------------------------------------------------------------
@@ -113,8 +108,10 @@ class Portal(object):
         self.sessions = PORTAL_SESSIONS
         self.sessions.portal = self
         self.process_id = os.getpid()
+
         self.server_process_id = None
         self.server_restart_mode = "shutdown"
+        self.server_info_dict = {}
 
         # set a callback if the server is killed abruptly,
         # by Ctrl-C, reboot etc.
@@ -163,9 +160,7 @@ class Portal(object):
             return
         self.sessions.disconnect_all()
         self.set_restart_mode(restart)
-        if os.name == 'nt' and os.path.exists(PORTAL_PIDFILE):
+
-            # for Windows we need to remove pid files manually
-            os.remove(PORTAL_PIDFILE)
         if not _reactor_stopping:
             # shutting down the reactor will trigger another signal. We set
             # a flag to avoid loops.

evennia/server/server.py

@@ -51,9 +51,9 @@ SERVER_STARTSTOP_MODULE = mod_import(settings.AT_SERVER_STARTSTOP_MODULE)
 SERVER_SERVICES_PLUGIN_MODULES = [mod_import(module) for module in make_iter(settings.SERVER_SERVICES_PLUGIN_MODULES)]
 
 
-#------------------------------------------------------------
+# ------------------------------------------------------------
 # Evennia Server settings
-#------------------------------------------------------------
+# ------------------------------------------------------------
 
 SERVERNAME = settings.SERVERNAME
 VERSION = get_evennia_version()