functional updates to ansible-connection (#18574)

* sends the serialized play_context into an already established connection
* hooks the alarm_handler() method in the connection plugin if it exists
* added configuration options for connect interval and retries
* adds syslog logging to Server() instance

This update will send the updated play_context back into an already
established connection in case privilege escalation / descalation activities
need to be performed.  This change will also hook the alarm_handler() method
in the connection instance (if available) and call it in case of a
sigalarm raised.

This update adds two new configuration options

* PERSISTENT_CONNECT_INTERVAL - time to wait in between connection attempts
* PERSISTENT_CONNECT_RETRIES - max number of retries

bin/ansible-connection

@@ -37,6 +37,8 @@ import struct
 import sys
 import time
 import traceback
+import syslog
+import datetime
 
 from io import BytesIO
 
@@ -47,6 +49,7 @@ from ansible.playbook.play_context import PlayContext
 from ansible.plugins import connection_loader
 from ansible.utils.path import unfrackpath, makedirs_safe
 
+
 def do_fork():
     '''
     Does the required double fork for a daemon process. Based on
@@ -97,25 +100,46 @@ def recv_data(s):
         data += d
     return data
 
+
 class Server():
+
     def __init__(self, path, play_context):
         self.path = path
         self.play_context = play_context
 
-        # FIXME: the connection loader here is created brand new,
+        self._start_time = datetime.datetime.now()
-        #        so it will not see any custom paths loaded (ie. via
+
-        #        roles), so we will need to serialize the connection
+        try:
-        #        loader and send it over as we do the PlayContext
+            # FIXME: the connection loader here is created brand new,
-        #        in the main() method below.
+            #        so it will not see any custom paths loaded (ie. via
-        self.conn = connection_loader.get(play_context.connection, play_context, sys.stdin)
+            #        roles), so we will need to serialize the connection
-        self.conn._connect()
+            #        loader and send it over as we do the PlayContext
+            #        in the main() method below.
+            self.log('loading connection plugin %s' % str(play_context.connection))
+            self.conn = connection_loader.get(play_context.connection, play_context, sys.stdin)
+            self.conn._connect()
+
+            self.socket = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
+            self.socket.bind(path)
+            self.socket.listen(1)
 
-        self.socket = socket.socket(socket.AF_UNIX, socket.SOCK_STREAM)
+        except Exception as exc:
-        self.socket.bind(path)
+            self.log(exc)
-        self.socket.listen(1)
+            return
 
         signal.signal(signal.SIGALRM, self.alarm_handler)
 
+    def log(self, msg):
+        syslog_msg = '[%s] %s' % (self.play_context.remote_addr, msg)
+        facility = getattr(syslog, C.DEFAULT_SYSLOG_FACILITY, syslog.LOG_USER)
+        syslog.openlog('ansible-connection', 0, facility)
+        syslog.syslog(syslog.LOG_INFO, syslog_msg)
+
+    def dispatch(self, obj, name, *args, **kwargs):
+        meth = getattr(obj, name, None)
+        if meth:
+            return meth(*args, **kwargs)
+
     def alarm_handler(self, signum, frame):
         '''
         Alarm handler
@@ -124,6 +148,9 @@ class Server():
         #        areas of code to check, so they can terminate
         #        earlier than the socket going back to the accept
         #        call and failing there.
+        #
+        # hooks the connection plugin to handle any cleanup
+        self.dispatch(self.conn, 'alarm_handler', signum, frame)
         self.socket.close()
 
     def run(self):
@@ -150,6 +177,8 @@ class Server():
                     if not data:
                         break
 
+                    signal.alarm(C.DEFAULT_TIMEOUT)
+
                     rc = 255
                     try:
                         if data.startswith(b'EXEC: '):
@@ -166,6 +195,18 @@ class Server():
                                 rc = 0
                             except:
                                 pass
+                        elif data.startswith(b'CONTEXT: '):
+                            pc_data = data.split(b'CONTEXT: ')[1]
+
+                            src = StringIO(pc_data)
+                            pc_data = cPickle.load(src)
+                            src.close()
+
+                            pc = PlayContext()
+                            pc.deserialize(pc_data)
+
+                            self.dispatch(self.conn, 'update_play_context', pc)
+                            continue
                         else:
                             stdout = ''
                             stderr = 'Invalid action specified'
@@ -173,19 +214,25 @@ class Server():
                         stdout = ''
                         stderr = traceback.format_exc()
 
+                    signal.alarm(0)
+
                     send_data(s, to_bytes(str(rc)))
                     send_data(s, to_bytes(stdout))
                     send_data(s, to_bytes(stderr))
                 s.close()
         except Exception as e:
             # FIXME: proper logging and error handling here
-            print("run exception: %s" % e)
+            self.log('runtime exception: %s' % e)
             print(traceback.format_exc())
         finally:
             # when done, close the connection properly and cleanup
             # the socket file so it can be recreated
+            end_time = datetime.datetime.now()
+            delta = end_time - self._start_time
+            self.log('shutting down connection, connection was active for %s secs' % delta)
             try:
                 self.conn.close()
+                self.socket.close()
             except Exception as e:
                 pass
             os.remove(self.path)
@@ -205,7 +252,7 @@ def main():
             cur_line = sys.stdin.readline()
         src = BytesIO(to_bytes(init_data))
         pc_data = cPickle.load(src)
-        src.close()
+        #src.close()
 
         pc = PlayContext()
         pc.deserialize(pc_data)
@@ -236,11 +283,11 @@ def main():
     if not os.path.exists(sf_path):
         pid = do_fork()
         if pid == 0:
-           server = Server(sf_path, pc)
+            server = Server(sf_path, pc)
-           fcntl.lockf(lock_fd, fcntl.LOCK_UN)
+            fcntl.lockf(lock_fd, fcntl.LOCK_UN)
-           os.close(lock_fd)
+            os.close(lock_fd)
-           server.run()
+            server.run()
-           sys.exit(0)
+            sys.exit(0)
     fcntl.lockf(lock_fd, fcntl.LOCK_UN)
     os.close(lock_fd)
 
@@ -262,24 +309,32 @@ def main():
                 break
             except socket.error:
                 # FIXME: better error handling/logging/message here
-                # FIXME: make # of retries configurable?
+                time.sleep(C.PERSISTENT_CONNECT_INTERVAL)
-                time.sleep(0.1)
                 attempts += 1
-                if attempts > 10:
+                if attempts > C.PERSISTENT_CONNECT_RETRIES:
-                    sys.stderr.write("failed to connect to the host, connection timeout\n")
+                    sys.stderr.write("failed to connect to the host, connection timeout")
                     sys.exit(255)
 
+        #
+        # send the play_context back into the connection so the connection
+        # can handle any privilege escalation or deescalation activities
+        #
+        pc_data = 'CONTEXT: %s' % src.getvalue()
+        send_data(sf, to_bytes(pc_data))
+        src.close()
+
         send_data(sf, to_bytes(data.strip()))
+
         rc = int(recv_data(sf), 10)
         stdout = recv_data(sf)
         stderr = recv_data(sf)
+
         sys.stdout.write(to_native(stdout))
         sys.stderr.write(to_native(stderr))
-        #sys.stdout.flush()
-        #sys.stderr.flush()
 
         sf.close()
         break
+
     sys.exit(rc)
 
 if __name__ == '__main__':

lib/ansible/constants.py

@@ -308,6 +308,8 @@ PARAMIKO_HOST_KEY_AUTO_ADD     = get_config(p, 'paramiko_connection', 'host_key_
 PARAMIKO_PROXY_COMMAND         = get_config(p, 'paramiko_connection', 'proxy_command', 'ANSIBLE_PARAMIKO_PROXY_COMMAND', None)
 PARAMIKO_LOOK_FOR_KEYS         = get_config(p, 'paramiko_connection', 'look_for_keys', 'ANSIBLE_PARAMIKO_LOOK_FOR_KEYS', True, value_type='boolean')
 PERSISTENT_CONNECT_TIMEOUT     = get_config(p, 'persistent_connection', 'connect_timeout', 'ANSIBLE_PERSISTENT_CONNECT_TIMEOUT', 30, value_type='integer')
+PERSISTENT_CONNECT_RETRIES     = get_config(p, 'persistent_connection', 'connect_retries', 'ANSIBLE_PERSISTENT_CONNECT_RETRIES', 10, value_type='integer')
+PERSISTENT_CONNECT_INTERVAL    = get_config(p, 'persistent_connection', 'connect_interval', 'ANSIBLE_PERSISTENT_CONNECT_INTERVAL', 1, value_type='integer')
 
 # obsolete -- will be formally removed
 ZEROMQ_PORT                    = get_config(p, 'fireball_connection', 'zeromq_port', 'ANSIBLE_ZEROMQ_PORT', 5099, value_type='integer')