# Copyright 2014-present MongoDB, Inc. # # Licensed under the Apache License, Version 2.0 (the "License"); you # may not use this file except in compliance with the License. You # may obtain a copy of the License at # # http://www.apache.org/licenses/LICENSE-2.0 # # Unless required by applicable law or agreed to in writing, software # distributed under the License is distributed on an "AS IS" BASIS, # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or # implied. See the License for the specific language governing # permissions and limitations under the License. """Internal class to monitor a topology of one or more servers.""" import os import random import threading import warnings import weakref from bson.py3compat import PY3, itervalues if PY3: import queue as Queue else: import Queue from pymongo import common, helpers, periodic_executor from pymongo.client_session import _ServerSessionPool from pymongo.errors import ( ConfigurationError, ConnectionFailure, NetworkTimeout, NotPrimaryError, OperationFailure, ServerSelectionTimeoutError, WriteError, ) from pymongo.ismaster import IsMaster from pymongo.monitor import SrvMonitor from pymongo.monotonic import time as _time from pymongo.pool import PoolOptions from pymongo.server import Server from pymongo.server_description import ServerDescription from pymongo.server_selectors import ( Selection, any_server_selector, arbiter_server_selector, readable_server_selector, secondary_server_selector, writable_server_selector, ) from pymongo.topology_description import ( SRV_POLLING_TOPOLOGIES, TOPOLOGY_TYPE, TopologyDescription, _updated_topology_description_srv_polling, updated_topology_description, ) def process_events_queue(queue_ref): q = queue_ref() if not q: return False # Cancel PeriodicExecutor. while True: try: event = q.get_nowait() except Queue.Empty: break else: fn, args = event fn(*args) return True # Continue PeriodicExecutor. class Topology(object): """Monitor a topology of one or more servers.""" def __init__(self, topology_settings): self._topology_id = topology_settings._topology_id self._listeners = topology_settings._pool_options._event_listeners pub = self._listeners is not None self._publish_server = pub and self._listeners.enabled_for_server self._publish_tp = pub and self._listeners.enabled_for_topology # Create events queue if there are publishers. self._events = None self.__events_executor = None if self._publish_server or self._publish_tp: self._events = Queue.Queue(maxsize=100) if self._publish_tp: self._events.put((self._listeners.publish_topology_opened, (self._topology_id,))) self._settings = topology_settings topology_description = TopologyDescription( topology_settings.get_topology_type(), topology_settings.get_server_descriptions(), topology_settings.replica_set_name, None, None, topology_settings, ) self._description = topology_description if self._publish_tp: initial_td = TopologyDescription( TOPOLOGY_TYPE.Unknown, {}, None, None, None, self._settings ) self._events.put( ( self._listeners.publish_topology_description_changed, (initial_td, self._description, self._topology_id), ) ) for seed in topology_settings.seeds: if self._publish_server: self._events.put((self._listeners.publish_server_opened, (seed, self._topology_id))) # Store the seed list to help diagnose errors in _error_message(). self._seed_addresses = list(topology_description.server_descriptions()) self._opened = False self._lock = threading.Lock() self._condition = self._settings.condition_class(self._lock) self._servers = {} self._pid = None self._max_cluster_time = None self._session_pool = _ServerSessionPool() if self._publish_server or self._publish_tp: def target(): return process_events_queue(weak) executor = periodic_executor.PeriodicExecutor( interval=common.EVENTS_QUEUE_FREQUENCY, min_interval=0.5, target=target, name="pymongo_events_thread", ) # We strongly reference the executor and it weakly references # the queue via this closure. When the topology is freed, stop # the executor soon. weak = weakref.ref(self._events, executor.close) self.__events_executor = executor executor.open() self._srv_monitor = None if self._settings.fqdn is not None and not self._settings.load_balanced: self._srv_monitor = SrvMonitor(self, self._settings) def open(self): """Start monitoring, or restart after a fork. No effect if called multiple times. .. warning:: Topology is shared among multiple threads and is protected by mutual exclusion. Using Topology from a process other than the one that initialized it will emit a warning and may result in deadlock. To prevent this from happening, MongoClient must be created after any forking. """ if self._pid is None: self._pid = os.getpid() else: if os.getpid() != self._pid: warnings.warn( "MongoClient opened before fork. Create MongoClient only " "after forking. See PyMongo's documentation for details: " "https://pymongo.readthedocs.io/en/stable/faq.html#" "is-pymongo-fork-safe" ) with self._lock: # Reset the session pool to avoid duplicate sessions in # the child process. self._session_pool.reset() with self._lock: self._ensure_opened() def select_servers(self, selector, server_selection_timeout=None, address=None): """Return a list of Servers matching selector, or time out. :Parameters: - `selector`: function that takes a list of Servers and returns a subset of them. - `server_selection_timeout` (optional): maximum seconds to wait. If not provided, the default value common.SERVER_SELECTION_TIMEOUT is used. - `address`: optional server address to select. Calls self.open() if needed. Raises exc:`ServerSelectionTimeoutError` after `server_selection_timeout` if no matching servers are found. """ if server_selection_timeout is None: server_timeout = self._settings.server_selection_timeout else: server_timeout = server_selection_timeout with self._lock: server_descriptions = self._select_servers_loop(selector, server_timeout, address) return [self.get_server_by_address(sd.address) for sd in server_descriptions] def _select_servers_loop(self, selector, timeout, address): """select_servers() guts. Hold the lock when calling this.""" now = _time() end_time = now + timeout server_descriptions = self._description.apply_selector( selector, address, custom_selector=self._settings.server_selector ) while not server_descriptions: # No suitable servers. if timeout == 0 or now > end_time: raise ServerSelectionTimeoutError( "%s, Timeout: %ss, Topology Description: %r" % (self._error_message(selector), timeout, self.description) ) self._ensure_opened() self._request_check_all() # Release the lock and wait for the topology description to # change, or for a timeout. We won't miss any changes that # came after our most recent apply_selector call, since we've # held the lock until now. self._condition.wait(common.MIN_HEARTBEAT_INTERVAL) self._description.check_compatible() now = _time() server_descriptions = self._description.apply_selector( selector, address, custom_selector=self._settings.server_selector ) self._description.check_compatible() return server_descriptions def select_server(self, selector, server_selection_timeout=None, address=None): """Like select_servers, but choose a random server if several match.""" return random.choice(self.select_servers(selector, server_selection_timeout, address)) def select_server_by_address(self, address, server_selection_timeout=None): """Return a Server for "address", reconnecting if necessary. If the server's type is not known, request an immediate check of all servers. Time out after "server_selection_timeout" if the server cannot be reached. :Parameters: - `address`: A (host, port) pair. - `server_selection_timeout` (optional): maximum seconds to wait. If not provided, the default value common.SERVER_SELECTION_TIMEOUT is used. Calls self.open() if needed. Raises exc:`ServerSelectionTimeoutError` after `server_selection_timeout` if no matching servers are found. """ return self.select_server(any_server_selector, server_selection_timeout, address) def _process_change(self, server_description, reset_pool=False): """Process a new ServerDescription on an opened topology. Hold the lock when calling this. """ td_old = self._description sd_old = td_old._server_descriptions[server_description.address] if _is_stale_server_description(sd_old, server_description): # This is a stale hello response. Ignore it. return suppress_event = (self._publish_server or self._publish_tp) and sd_old == server_description if self._publish_server and not suppress_event: self._events.put( ( self._listeners.publish_server_description_changed, (sd_old, server_description, server_description.address, self._topology_id), ) ) self._description = updated_topology_description(self._description, server_description) self._update_servers() self._receive_cluster_time_no_lock(server_description.cluster_time) if self._publish_tp and not suppress_event: self._events.put( ( self._listeners.publish_topology_description_changed, (td_old, self._description, self._topology_id), ) ) # Shutdown SRV polling for unsupported cluster types. # This is only applicable if the old topology was Unknown, and the # new one is something other than Unknown or Sharded. if self._srv_monitor and ( td_old.topology_type == TOPOLOGY_TYPE.Unknown and self._description.topology_type not in SRV_POLLING_TOPOLOGIES ): self._srv_monitor.close() # Clear the pool from a failed heartbeat. if reset_pool: server = self._servers.get(server_description.address) if server: server.pool.reset() # Wake waiters in select_servers(). self._condition.notify_all() def on_change(self, server_description, reset_pool=False): """Process a new ServerDescription after a hello call completes.""" # We do no I/O holding the lock. with self._lock: # Monitors may continue working on hello calls for some time # after a call to Topology.close, so this method may be called at # any time. Ensure the topology is open before processing the # change. # Any monitored server was definitely in the topology description # once. Check if it's still in the description or if some state- # change removed it. E.g., we got a host list from the primary # that didn't include this server. if self._opened and self._description.has_server(server_description.address): self._process_change(server_description, reset_pool) def _process_srv_update(self, seedlist): """Process a new seedlist on an opened topology. Hold the lock when calling this. """ td_old = self._description if td_old.topology_type not in SRV_POLLING_TOPOLOGIES: return self._description = _updated_topology_description_srv_polling(self._description, seedlist) self._update_servers() if self._publish_tp: self._events.put( ( self._listeners.publish_topology_description_changed, (td_old, self._description, self._topology_id), ) ) def on_srv_update(self, seedlist): """Process a new list of nodes obtained from scanning SRV records.""" # We do no I/O holding the lock. with self._lock: if self._opened: self._process_srv_update(seedlist) def get_server_by_address(self, address): """Get a Server or None. Returns the current version of the server immediately, even if it's Unknown or absent from the topology. Only use this in unittests. In driver code, use select_server_by_address, since then you're assured a recent view of the server's type and wire protocol version. """ return self._servers.get(address) def has_server(self, address): return address in self._servers def get_primary(self): """Return primary's address or None.""" # Implemented here in Topology instead of MongoClient, so it can lock. with self._lock: topology_type = self._description.topology_type if topology_type != TOPOLOGY_TYPE.ReplicaSetWithPrimary: return None return writable_server_selector(self._new_selection())[0].address def _get_replica_set_members(self, selector): """Return set of replica set member addresses.""" # Implemented here in Topology instead of MongoClient, so it can lock. with self._lock: topology_type = self._description.topology_type if topology_type not in ( TOPOLOGY_TYPE.ReplicaSetWithPrimary, TOPOLOGY_TYPE.ReplicaSetNoPrimary, ): return set() return set([sd.address for sd in selector(self._new_selection())]) def get_secondaries(self): """Return set of secondary addresses.""" return self._get_replica_set_members(secondary_server_selector) def get_arbiters(self): """Return set of arbiter addresses.""" return self._get_replica_set_members(arbiter_server_selector) def max_cluster_time(self): """Return a document, the highest seen $clusterTime.""" return self._max_cluster_time def _receive_cluster_time_no_lock(self, cluster_time): # Driver Sessions Spec: "Whenever a driver receives a cluster time from # a server it MUST compare it to the current highest seen cluster time # for the deployment. If the new cluster time is higher than the # highest seen cluster time it MUST become the new highest seen cluster # time. Two cluster times are compared using only the BsonTimestamp # value of the clusterTime embedded field." if cluster_time: # ">" uses bson.timestamp.Timestamp's comparison operator. if ( not self._max_cluster_time or cluster_time["clusterTime"] > self._max_cluster_time["clusterTime"] ): self._max_cluster_time = cluster_time def receive_cluster_time(self, cluster_time): with self._lock: self._receive_cluster_time_no_lock(cluster_time) def request_check_all(self, wait_time=5): """Wake all monitors, wait for at least one to check its server.""" with self._lock: self._request_check_all() self._condition.wait(wait_time) def handle_getlasterror(self, address, error_msg): """Clear our pool for a server, mark it Unknown, and check it soon.""" error = NotPrimaryError(error_msg, {"code": 10107, "errmsg": error_msg}) with self._lock: server = self._servers.get(address) if server: self._process_change(ServerDescription(address, error=error), True) server.request_check() def data_bearing_servers(self): """Return a list of all data-bearing servers. This includes any server that might be selected for an operation. """ if self._description.topology_type == TOPOLOGY_TYPE.Single: return self._description.known_servers return self._description.readable_servers def update_pool(self, all_credentials): # Remove any stale sockets and add new sockets if pool is too small. servers = [] with self._lock: # Only update pools for data-bearing servers. for sd in self.data_bearing_servers(): server = self._servers[sd.address] servers.append((server, server.pool.gen.get_overall())) for server, generation in servers: server.pool.remove_stale_sockets(generation, all_credentials) def close(self): """Clear pools and terminate monitors. Topology reopens on demand.""" with self._lock: for server in self._servers.values(): server.close() # Mark all servers Unknown. self._description = self._description.reset() for address, sd in self._description.server_descriptions().items(): if address in self._servers: self._servers[address].description = sd # Stop SRV polling thread. if self._srv_monitor: self._srv_monitor.close() self._opened = False # Publish only after releasing the lock. if self._publish_tp: self._events.put((self._listeners.publish_topology_closed, (self._topology_id,))) if self._publish_server or self._publish_tp: self.__events_executor.close() @property def description(self): return self._description def pop_all_sessions(self): """Pop all session ids from the pool.""" with self._lock: return self._session_pool.pop_all() def _check_session_support(self): """Internal check for session support on non-load balanced clusters.""" session_timeout = self._description.logical_session_timeout_minutes if session_timeout is None: # Maybe we need an initial scan? Can raise ServerSelectionError. if self._description.topology_type == TOPOLOGY_TYPE.Single: if not self._description.has_known_servers: self._select_servers_loop( any_server_selector, self._settings.server_selection_timeout, None ) elif not self._description.readable_servers: self._select_servers_loop( readable_server_selector, self._settings.server_selection_timeout, None ) session_timeout = self._description.logical_session_timeout_minutes if session_timeout is None: raise ConfigurationError("Sessions are not supported by this MongoDB deployment") return session_timeout def get_server_session(self): """Start or resume a server session, or raise ConfigurationError.""" with self._lock: # Sessions are always supported in load balanced mode. if not self._settings.load_balanced: session_timeout = self._check_session_support() else: # Sessions never time out in load balanced mode. session_timeout = float("inf") return self._session_pool.get_server_session(session_timeout) def return_server_session(self, server_session, lock): if lock: with self._lock: self._session_pool.return_server_session( server_session, self._description.logical_session_timeout_minutes ) else: # Called from a __del__ method, can't use a lock. self._session_pool.return_server_session_no_lock(server_session) def _new_selection(self): """A Selection object, initially including all known servers. Hold the lock when calling this. """ return Selection.from_topology_description(self._description) def _ensure_opened(self): """Start monitors, or restart after a fork. Hold the lock when calling this. """ if not self._opened: self._opened = True self._update_servers() # Start or restart the events publishing thread. if self._publish_tp or self._publish_server: self.__events_executor.open() # Start the SRV polling thread. if self._srv_monitor and (self.description.topology_type in SRV_POLLING_TOPOLOGIES): self._srv_monitor.open() if self._settings.load_balanced: # Emit initial SDAM events for load balancer mode. self._process_change( ServerDescription( self._seed_addresses[0], IsMaster({"ok": 1, "serviceId": self._topology_id, "maxWireVersion": 13}), ) ) # Ensure that the monitors are open. for server in itervalues(self._servers): server.open() def _is_stale_error(self, address, err_ctx): server = self._servers.get(address) if server is None: # Another thread removed this server from the topology. return True if server._pool.stale_generation(err_ctx.sock_generation, err_ctx.service_id): # This is an outdated error from a previous pool version. return True # topologyVersion check, ignore error when cur_tv >= error_tv: cur_tv = server.description.topology_version error = err_ctx.error error_tv = None if error and hasattr(error, "details"): if isinstance(error.details, dict): error_tv = error.details.get("topologyVersion") return _is_stale_error_topology_version(cur_tv, error_tv) def _handle_error(self, address, err_ctx): if self._is_stale_error(address, err_ctx): return server = self._servers[address] error = err_ctx.error exc_type = type(error) service_id = err_ctx.service_id if issubclass(exc_type, NetworkTimeout) and err_ctx.completed_handshake: # The socket has been closed. Don't reset the server. # Server Discovery And Monitoring Spec: "When an application # operation fails because of any network error besides a socket # timeout...." return elif issubclass(exc_type, WriteError): # Ignore writeErrors. return elif issubclass(exc_type, NotPrimaryError): # As per the SDAM spec if: # - the server sees a "not primary" error, and # - the server is not shutting down, and # - the server version is >= 4.2, then # we keep the existing connection pool, but mark the server type # as Unknown and request an immediate check of the server. # Otherwise, we clear the connection pool, mark the server as # Unknown and request an immediate check of the server. err_code = error.details.get("code", -1) is_shutting_down = err_code in helpers._SHUTDOWN_CODES # Mark server Unknown, clear the pool, and request check. if not self._settings.load_balanced: self._process_change(ServerDescription(address, error=error)) if is_shutting_down or (err_ctx.max_wire_version <= 7): # Clear the pool. server.reset(service_id) server.request_check() elif issubclass(exc_type, ConnectionFailure): # "Client MUST replace the server's description with type Unknown # ... MUST NOT request an immediate check of the server." if not self._settings.load_balanced: self._process_change(ServerDescription(address, error=error)) # Clear the pool. server.reset(service_id) # "When a client marks a server Unknown from `Network error when # reading or writing`_, clients MUST cancel the hello check on # that server and close the current monitoring connection." server._monitor.cancel_check() elif issubclass(exc_type, OperationFailure): # Do not request an immediate check since the server is likely # shutting down. if error.code in helpers._NOT_MASTER_CODES: if not self._settings.load_balanced: self._process_change(ServerDescription(address, error=error)) # Clear the pool. server.reset(service_id) def handle_error(self, address, err_ctx): """Handle an application error. May reset the server to Unknown, clear the pool, and request an immediate check depending on the error and the context. """ with self._lock: self._handle_error(address, err_ctx) def _request_check_all(self): """Wake all monitors. Hold the lock when calling this.""" for server in self._servers.values(): server.request_check() def _update_servers(self): """Sync our Servers from TopologyDescription.server_descriptions. Hold the lock while calling this. """ for address, sd in self._description.server_descriptions().items(): if address not in self._servers: monitor = self._settings.monitor_class( server_description=sd, topology=self, pool=self._create_pool_for_monitor(address), topology_settings=self._settings, ) weak = None if self._publish_server: weak = weakref.ref(self._events) server = Server( server_description=sd, pool=self._create_pool_for_server(address), monitor=monitor, topology_id=self._topology_id, listeners=self._listeners, events=weak, ) self._servers[address] = server server.open() else: # Cache old is_writable value. was_writable = self._servers[address].description.is_writable # Update server description. self._servers[address].description = sd # Update is_writable value of the pool, if it changed. if was_writable != sd.is_writable: self._servers[address].pool.update_is_writable(sd.is_writable) for address, server in list(self._servers.items()): if not self._description.has_server(address): server.close() self._servers.pop(address) def _create_pool_for_server(self, address): return self._settings.pool_class(address, self._settings.pool_options) def _create_pool_for_monitor(self, address): options = self._settings.pool_options # According to the Server Discovery And Monitoring Spec, monitors use # connect_timeout for both connect_timeout and socket_timeout. The # pool only has one socket so maxPoolSize and so on aren't needed. monitor_pool_options = PoolOptions( connect_timeout=options.connect_timeout, socket_timeout=options.connect_timeout, ssl_context=options._ssl_context, ssl_match_hostname=options.ssl_match_hostname, event_listeners=options._event_listeners, appname=options.appname, driver=options.driver, server_api=options.server_api, ) return self._settings.pool_class(address, monitor_pool_options, handshake=False) def _error_message(self, selector): """Format an error message if server selection fails. Hold the lock when calling this. """ is_replica_set = self._description.topology_type in ( TOPOLOGY_TYPE.ReplicaSetWithPrimary, TOPOLOGY_TYPE.ReplicaSetNoPrimary, ) if is_replica_set: server_plural = "replica set members" elif self._description.topology_type == TOPOLOGY_TYPE.Sharded: server_plural = "mongoses" else: server_plural = "servers" if self._description.known_servers: # We've connected, but no servers match the selector. if selector is writable_server_selector: if is_replica_set: return "No primary available for writes" else: return "No %s available for writes" % server_plural else: return 'No %s match selector "%s"' % (server_plural, selector) else: addresses = list(self._description.server_descriptions()) servers = list(self._description.server_descriptions().values()) if not servers: if is_replica_set: # We removed all servers because of the wrong setName? return 'No %s available for replica set name "%s"' % ( server_plural, self._settings.replica_set_name, ) else: return "No %s available" % server_plural # 1 or more servers, all Unknown. Are they unknown for one reason? error = servers[0].error same = all(server.error == error for server in servers[1:]) if same: if error is None: # We're still discovering. return "No %s found yet" % server_plural if is_replica_set and not set(addresses).intersection(self._seed_addresses): # We replaced our seeds with new hosts but can't reach any. return ( "Could not reach any servers in %s. Replica set is" " configured with internal hostnames or IPs?" % addresses ) return str(error) else: return ",".join(str(server.error) for server in servers if server.error) def __repr__(self): msg = "" if not self._opened: msg = "CLOSED " return "<%s %s%r>" % (self.__class__.__name__, msg, self._description) class _ErrorContext(object): """An error with context for SDAM error handling.""" def __init__(self, error, max_wire_version, sock_generation, completed_handshake, service_id): self.error = error self.max_wire_version = max_wire_version self.sock_generation = sock_generation self.completed_handshake = completed_handshake self.service_id = service_id def _is_stale_error_topology_version(current_tv, error_tv): """Return True if the error's topologyVersion is <= current.""" if current_tv is None or error_tv is None: return False if current_tv["processId"] != error_tv["processId"]: return False return current_tv["counter"] >= error_tv["counter"] def _is_stale_server_description(current_sd, new_sd): """Return True if the new topologyVersion is < current.""" current_tv, new_tv = current_sd.topology_version, new_sd.topology_version if current_tv is None or new_tv is None: return False if current_tv["processId"] != new_tv["processId"]: return False return current_tv["counter"] > new_tv["counter"]