import asyncio
import contextlib
import copy
import inspect
import math
import socket
import sys
import time
import warnings
import weakref
from abc import ABC, abstractmethod
from itertools import chain
from types import MappingProxyType
from typing import (
Any,
AsyncIterator,
Callable,
Iterable,
List,
Literal,
Mapping,
Optional,
Protocol,
Set,
Tuple,
Type,
TypedDict,
TypeVar,
Union,
)
from urllib.parse import ParseResult, parse_qs, unquote, urlparse
from ..observability.attributes import (
DB_CLIENT_CONNECTION_POOL_NAME,
DB_CLIENT_CONNECTION_STATE,
AttributeBuilder,
ConnectionState,
get_pool_name,
)
from ..utils import SSL_AVAILABLE, deprecated_function
if SSL_AVAILABLE:
import ssl
from ssl import SSLContext, TLSVersion, VerifyFlags
else:
ssl = None
TLSVersion = None
SSLContext = None
VerifyFlags = None
from ..auth.token import TokenInterface
from ..driver_info import DriverInfo, resolve_driver_info
from ..event import AsyncAfterConnectionReleasedEvent, EventDispatcher
from ..utils import deprecated_args, format_error_message
# the functionality is available in 3.11.x but has a major issue before
# 3.11.3. See https://github.com/redis/redis-py/issues/2633
if sys.version_info >= (3, 11, 3):
from asyncio import timeout as async_timeout
else:
from async_timeout import timeout as async_timeout
from redis.asyncio.maint_notifications import (
AsyncMaintNotificationsConnectionHandler,
AsyncMaintNotificationsPoolHandler,
AsyncOSSMaintNotificationsHandler,
)
from redis.asyncio.observability.recorder import (
record_connection_closed,
record_connection_count,
record_connection_create_time,
record_connection_wait_time,
record_error_count,
)
from redis.asyncio.retry import Retry
from redis.backoff import NoBackoff
from redis.credentials import CredentialProvider, UsernamePasswordCredentialProvider
from redis.exceptions import (
AuthenticationError,
AuthenticationWrongNumberOfArgsError,
ConnectionError,
DataError,
MaxConnectionsError,
RedisError,
ResponseError,
TimeoutError,
)
from redis.himport import HImportRegistry
from redis.maint_notifications import (
MaintenanceState,
MaintNotificationsConfig,
NodeMovingNotification,
_build_moving_cleanup_connection_kwargs,
_build_moving_connection_kwargs,
)
from redis.observability.metrics import CloseReason
from redis.typing import EncodableT
from redis.utils import (
DEFAULT_RESP_VERSION,
HIREDIS_AVAILABLE,
SENTINEL,
check_protocol_version,
str_if_bytes,
)
from .._defaults import (
DEFAULT_SOCKET_CONNECT_TIMEOUT,
DEFAULT_SOCKET_READ_SIZE,
DEFAULT_SOCKET_TIMEOUT,
get_default_socket_keepalive_options,
)
from .._parsers import (
AsyncPushNotificationsParser,
BaseParser,
Encoder,
_AsyncHiredisParser,
_AsyncRESP2Parser,
_AsyncRESP3Parser,
)
SYM_STAR = b"*"
SYM_DOLLAR = b"$"
SYM_CRLF = b"\r\n"
SYM_LF = b"\n"
SYM_EMPTY = b""
DefaultParser: Type[Union[_AsyncRESP2Parser, _AsyncRESP3Parser, _AsyncHiredisParser]]
if HIREDIS_AVAILABLE:
DefaultParser = _AsyncHiredisParser
else:
DefaultParser = _AsyncRESP3Parser
class ConnectCallbackProtocol(Protocol):
def __call__(self, connection: "AbstractConnection"): ...
class AsyncConnectCallbackProtocol(Protocol):
async def __call__(self, connection: "AbstractConnection"): ...
ConnectCallbackT = Union[ConnectCallbackProtocol, AsyncConnectCallbackProtocol]
class AsyncMaintNotificationsAbstractConnection:
"""
Internal mixin for async maintenance notification state and parser handlers.
The sync implementation uses the same mixin-style structure. The async
version keeps the notification state and parser handler installation close
to the connection without sending the server-side handshake; that is wired
in a later step.
"""
__slots__ = ()
def __init__(
self,
maint_notifications_config: MaintNotificationsConfig | None,
maint_notifications_pool_handler: (
AsyncMaintNotificationsPoolHandler | None
) = None,
maintenance_state: MaintenanceState = MaintenanceState.NONE,
maintenance_notification_hash: int | None = None,
orig_host_address: str | None = None,
orig_socket_timeout: float | None = None,
orig_socket_connect_timeout: float | None = None,
oss_cluster_maint_notifications_handler: (
AsyncOSSMaintNotificationsHandler | None
) = None,
parser: BaseParser | None = None,
) -> None:
self.maint_notifications_config = maint_notifications_config
self.maintenance_state = maintenance_state
self.maintenance_notification_hash = maintenance_notification_hash
self._processed_start_maint_notifications: set[int] = set()
self._skipped_end_maint_notifications: set[int] = set()
self._configure_maintenance_notifications(
maint_notifications_pool_handler,
orig_host_address,
orig_socket_timeout,
orig_socket_connect_timeout,
oss_cluster_maint_notifications_handler,
parser,
)
@abstractmethod
def _get_parser(self) -> BaseParser:
pass
def _get_push_notifications_parser(self) -> AsyncPushNotificationsParser:
parser = self._get_parser()
if not isinstance(parser, (_AsyncHiredisParser, _AsyncRESP3Parser)):
raise RedisError(
"Maintenance notifications are only supported with hiredis and RESP3 parsers!"
)
return parser
@abstractmethod
def get_protocol(self):
pass
@abstractmethod
async def send_command(self, *args: Any, **kwargs: Any) -> None:
pass
@abstractmethod
async def read_response(
self,
disable_decoding: bool = False,
timeout: float | None = None,
*,
disconnect_on_error: bool = True,
push_request: bool | None = False,
) -> Any:
pass
@abstractmethod
def getpeername(self) -> str | None:
pass
def _configure_maintenance_notifications(
self,
maint_notifications_pool_handler: (
AsyncMaintNotificationsPoolHandler | None
) = None,
orig_host_address: str | None = None,
orig_socket_timeout: float | None = None,
orig_socket_connect_timeout: float | None = None,
oss_cluster_maint_notifications_handler: (
AsyncOSSMaintNotificationsHandler | None
) = None,
parser: BaseParser | None = None,
) -> None:
if (
not self.maint_notifications_config
or not self.maint_notifications_config.enabled
):
self._maint_notifications_pool_handler = None
self._maint_notifications_connection_handler = None
self._oss_cluster_maint_notifications_handler = None
return
if not parser:
raise RedisError(
"To configure maintenance notifications, a parser must be provided!"
)
if not isinstance(parser, _AsyncHiredisParser) and not isinstance(
parser, _AsyncRESP3Parser
):
raise RedisError(
"Maintenance notifications are only supported with hiredis and RESP3 parsers!"
)
if maint_notifications_pool_handler:
# Extract a reference to a new pool handler that copies all properties
# of the original one and has a different connection reference
# This is needed because when we attach the handler to the parser
# we need to make sure that the handler has a reference to the
# connection that the parser is attached to.
self._maint_notifications_pool_handler = (
maint_notifications_pool_handler.get_handler_for_connection()
)
self._maint_notifications_pool_handler.set_connection(self)
else:
self._maint_notifications_pool_handler = None
self._maint_notifications_connection_handler = (
AsyncMaintNotificationsConnectionHandler(
self, self.maint_notifications_config
)
)
if oss_cluster_maint_notifications_handler:
self._oss_cluster_maint_notifications_handler = (
oss_cluster_maint_notifications_handler
)
parser.set_oss_cluster_maint_push_handler(
oss_cluster_maint_notifications_handler.handle_notification
)
else:
self._oss_cluster_maint_notifications_handler = None
# Set up pool handler to parser if available
if self._maint_notifications_pool_handler:
parser.set_node_moving_push_handler(
self._maint_notifications_pool_handler.handle_notification
)
# Set up connection handler
parser.set_maintenance_push_handler(
self._maint_notifications_connection_handler.handle_notification
)
self.orig_host_address = orig_host_address if orig_host_address else self.host
self.orig_socket_timeout = (
orig_socket_timeout if orig_socket_timeout else self.socket_timeout
)
self.orig_socket_connect_timeout = (
orig_socket_connect_timeout
if orig_socket_connect_timeout
else self.socket_connect_timeout
)
def set_maint_notifications_pool_handler_for_connection(
self, maint_notifications_pool_handler: AsyncMaintNotificationsPoolHandler
) -> None:
# Deep copy the pool handler to avoid sharing the same pool handler
# between multiple connections, because otherwise each connection will override
# the connection reference and the pool handler will only hold a reference
# to the last connection that was set.
maint_notifications_pool_handler_copy = (
maint_notifications_pool_handler.get_handler_for_connection()
)
maint_notifications_pool_handler_copy.set_connection(self)
parser = self._get_push_notifications_parser()
parser.set_node_moving_push_handler(
maint_notifications_pool_handler_copy.handle_notification
)
self._maint_notifications_pool_handler = maint_notifications_pool_handler_copy
# Update maintenance notification connection handler if it doesn't exist
if not self._maint_notifications_connection_handler:
self._maint_notifications_connection_handler = (
AsyncMaintNotificationsConnectionHandler(
self, maint_notifications_pool_handler.config
)
)
parser.set_maintenance_push_handler(
self._maint_notifications_connection_handler.handle_notification
)
else:
self._maint_notifications_connection_handler.config = (
maint_notifications_pool_handler.config
)
def set_maint_notifications_cluster_handler_for_connection(
self,
oss_cluster_maint_notifications_handler: AsyncOSSMaintNotificationsHandler,
) -> None:
parser = self._get_push_notifications_parser()
parser.set_oss_cluster_maint_push_handler(
oss_cluster_maint_notifications_handler.handle_notification
)
# OSS cluster mode and pool-handler mode are mutually exclusive. Clear
# any node-moving/pool handler a default (RESP3 "auto") pool wired in
# __init__ so this existing connection is not configured with both.
parser.set_node_moving_push_handler(None)
self._maint_notifications_pool_handler = None
self._oss_cluster_maint_notifications_handler = (
oss_cluster_maint_notifications_handler
)
# Update maintenance notification connection handler if it doesn't exist
if not self._maint_notifications_connection_handler:
self._maint_notifications_connection_handler = (
AsyncMaintNotificationsConnectionHandler(
self, oss_cluster_maint_notifications_handler.config
)
)
parser.set_maintenance_push_handler(
self._maint_notifications_connection_handler.handle_notification
)
else:
self._maint_notifications_connection_handler.config = (
oss_cluster_maint_notifications_handler.config
)
async def activate_maint_notifications_handling_if_enabled(
self, check_health: bool = True
) -> None:
# Send maintenance notifications handshake if RESP3 is active
# and maintenance notifications are enabled
# and we have a host to determine the endpoint type from
# When the maint_notifications_config enabled mode is "auto",
# we just log a warning if the handshake fails
# When the mode is enabled=True, we raise an exception in case of failure
host = getattr(self, "host", None)
if (
check_protocol_version(self.get_protocol(), 3)
and self.maint_notifications_config
and self.maint_notifications_config.enabled
and self._maint_notifications_connection_handler
and host is not None
):
await self._enable_maintenance_notifications(
maint_notifications_config=self.maint_notifications_config,
check_health=check_health,
)
async def _enable_maintenance_notifications(
self,
maint_notifications_config: MaintNotificationsConfig,
check_health: bool = True,
) -> None:
try:
host = getattr(self, "host", None)
if host is None:
raise ValueError(
"Cannot enable maintenance notifications for connection"
" object that doesn't have a host attribute."
)
endpoint_type = maint_notifications_config.get_endpoint_type(host, self)
await self.send_command(
"CLIENT",
"MAINT_NOTIFICATIONS",
"ON",
"moving-endpoint-type",
endpoint_type.value,
check_health=check_health,
)
response = await self.read_response()
if not response or str_if_bytes(response) != "OK":
raise ResponseError(
"The server doesn't support maintenance notifications"
)
except Exception as e:
if (
isinstance(e, ResponseError)
and maint_notifications_config.enabled == "auto"
):
# Log warning but don't fail the connection
import logging
logger = logging.getLogger(__name__)
logger.debug(f"Failed to enable maintenance notifications: {e}")
else:
raise
def get_resolved_ip(self) -> str | None:
"""
Extract the resolved IP address from an established connection or host.
First tries to get the actual peer IP from the async stream writer, then
falls back to DNS resolution if needed.
Returns:
The resolved IP address, or None if it cannot be determined.
"""
# Method 1: Try to get the actual IP from the established stream.
# This is most accurate as it shows the exact IP being used.
try:
peer_addr = self.getpeername()
if peer_addr:
return peer_addr
except (AttributeError, OSError):
# Stream might not be connected or peer address lookup might fail.
pass
# Method 2: Fall back to the configured host (which may be an IP or an
# FQDN). Unlike the sync client we intentionally do NOT call
# socket.getaddrinfo() here: this method runs on the event loop, so a
# blocking DNS resolution would stall it. On the endpoint-type handshake
# path (get_endpoint_type) getpeername() above always succeeds because the
# writer was just connected, so this fallback is only reached by the
# debug-log call sites during reconnects — where returning the host is
# fine. A blocking getaddrinfo on an FQDN host there can freeze the loop
# for seconds and trip unrelated connect timeouts.
return getattr(self, "host", None)
@property
def maintenance_state(self) -> MaintenanceState:
return self._maintenance_state
@maintenance_state.setter
def maintenance_state(self, state: MaintenanceState) -> None:
self._maintenance_state = state
def add_maint_start_notification(self, id: int) -> None:
self._processed_start_maint_notifications.add(id)
def get_processed_start_notifications(self) -> set[int]:
return self._processed_start_maint_notifications
def add_skipped_end_notification(self, id: int) -> None:
self._skipped_end_maint_notifications.add(id)
def get_skipped_end_notifications(self) -> set[int]:
return self._skipped_end_maint_notifications
def reset_received_notifications(self) -> None:
self._processed_start_maint_notifications.clear()
self._skipped_end_maint_notifications.clear()
def update_current_socket_timeout(
self, relaxed_timeout: float | None = None
) -> None:
timeout = relaxed_timeout if relaxed_timeout != -1 else self.socket_timeout
self._reschedule_active_read_timeout(timeout)
def _reschedule_active_read_timeout(self, timeout: float | None) -> None:
timeout_context = getattr(self, "_active_read_timeout", None)
if timeout_context is None:
# No read_response call is currently inside its socket timeout
# context, so there is no in-flight deadline to relax or restore.
return
if timeout is None:
# A None socket timeout means the active read should become blocking.
# Python 3.11's timeout context supports clearing the deadline.
if hasattr(timeout_context, "reschedule"):
timeout_context.reschedule(None)
# Older async-timeout contexts cannot clear a deadline, so reject the
# current timeout instead of leaving a stale relaxed deadline active.
elif hasattr(timeout_context, "reject"):
timeout_context.reject()
return
# Active read timeouts are stored as loop-time deadlines, not durations.
deadline = asyncio.get_running_loop().time() + timeout
if hasattr(timeout_context, "reschedule"):
# Python 3.11 asyncio.timeout exposes reschedule().
timeout_context.reschedule(deadline)
elif hasattr(timeout_context, "update"):
# async-timeout exposes update() for the same deadline adjustment.
timeout_context.update(deadline)
def set_tmp_settings(
self,
tmp_host_address: str | object | None = SENTINEL,
tmp_relaxed_timeout: float | None = -1,
) -> None:
"""
SENTINEL keeps the host unchanged. -1 keeps the relaxed timeout unchanged.
"""
if tmp_host_address and tmp_host_address != SENTINEL:
self.host = str(tmp_host_address)
if tmp_relaxed_timeout != -1:
self.socket_timeout = tmp_relaxed_timeout
self.socket_connect_timeout = tmp_relaxed_timeout
def reset_tmp_settings(
self,
reset_host_address: bool = False,
reset_relaxed_timeout: bool = False,
) -> None:
if reset_host_address:
self.host = self.orig_host_address
if reset_relaxed_timeout:
self.socket_timeout = self.orig_socket_timeout
self.socket_connect_timeout = self.orig_socket_connect_timeout
class AbstractConnection(AsyncMaintNotificationsAbstractConnection):
"""Manages communication to and from a Redis server"""
__slots__ = (
"db",
"username",
"client_name",
"lib_name",
"lib_version",
"credential_provider",
"password",
"socket_timeout",
"socket_connect_timeout",
"redis_connect_func",
"retry_on_timeout",
"retry_on_error",
"health_check_interval",
"next_health_check",
"last_active_at",
"encoder",
"ssl_context",
"protocol",
"_reader",
"_writer",
"_parser",
"_active_read_timeout",
"_connect_callbacks",
"_buffer_cutoff",
"_lock",
"_socket_read_size",
"__dict__",
)
@deprecated_args(
args_to_warn=["lib_name", "lib_version"],
reason="Use 'driver_info' parameter instead. "
"lib_name and lib_version will be removed in a future version.",
)
def __init__(
self,
*,
db: str | int = 0,
password: str | None = None,
socket_timeout: float | None = DEFAULT_SOCKET_TIMEOUT,
socket_connect_timeout: float | None = DEFAULT_SOCKET_CONNECT_TIMEOUT,
retry_on_timeout: bool = False,
retry_on_error: list | object = SENTINEL,
encoding: str = "utf-8",
encoding_errors: str = "strict",
decode_responses: bool = False,
parser_class: Type[BaseParser] = DefaultParser,
socket_read_size: int = DEFAULT_SOCKET_READ_SIZE,
health_check_interval: float = 0,
client_name: str | None = None,
lib_name: str | object | None = SENTINEL,
lib_version: str | object | None = SENTINEL,
driver_info: DriverInfo | object | None = SENTINEL,
username: str | None = None,
retry: Retry | None = None,
redis_connect_func: ConnectCallbackT | None = None,
encoder_class: Type[Encoder] = Encoder,
credential_provider: CredentialProvider | None = None,
protocol: int | None = None,
legacy_responses: bool = True,
event_dispatcher: EventDispatcher | None = None,
maint_notifications_config: MaintNotificationsConfig | None = None,
maint_notifications_pool_handler: (
AsyncMaintNotificationsPoolHandler | None
) = None,
maintenance_state: MaintenanceState = MaintenanceState.NONE,
maintenance_notification_hash: int | None = None,
orig_host_address: str | None = None,
orig_socket_timeout: float | None = None,
orig_socket_connect_timeout: float | None = None,
oss_cluster_maint_notifications_handler: (
AsyncOSSMaintNotificationsHandler | None
) = None,
himport_registry: HImportRegistry | None = None,
):
"""
Initialize a new async Connection.
Parameters
----------
driver_info : DriverInfo, optional
Driver metadata for CLIENT SETINFO. If provided, lib_name and lib_version
are ignored. If not provided, a DriverInfo will be created from lib_name
and lib_version. Explicit None disables CLIENT SETINFO.
lib_name : str, optional
**Deprecated.** Use driver_info instead. Library name for CLIENT SETINFO.
lib_version : str, optional
**Deprecated.** Use driver_info instead. Library version for CLIENT SETINFO.
"""
if (username or password) and credential_provider is not None:
raise DataError(
"'username' and 'password' cannot be passed along with 'credential_"
"provider'. Please provide only one of the following arguments: \n"
"1. 'password' and (optional) 'username'\n"
"2. 'credential_provider'"
)
if event_dispatcher is None:
self._event_dispatcher = EventDispatcher()
else:
self._event_dispatcher = event_dispatcher
self.db = db
self.client_name = client_name
# Handle driver_info: if provided, use it; otherwise create from lib_name/lib_version.
self.driver_info = resolve_driver_info(driver_info, lib_name, lib_version)
self.credential_provider = credential_provider
self.password = password
self.username = username
self.socket_timeout = socket_timeout
if socket_connect_timeout is None:
socket_connect_timeout = socket_timeout
self.socket_connect_timeout = socket_connect_timeout
self.retry_on_timeout = retry_on_timeout
if retry_on_error is SENTINEL:
retry_on_error = []
if retry_on_timeout:
retry_on_error.append(TimeoutError)
retry_on_error.append(socket.timeout)
retry_on_error.append(asyncio.TimeoutError)
self.retry_on_error = retry_on_error
if retry or retry_on_error:
if not retry:
self.retry = Retry(NoBackoff(), 1)
else:
# deep-copy the Retry object as it is mutable
self.retry = copy.deepcopy(retry)
# Update the retry's supported errors with the specified errors
self.retry.update_supported_errors(retry_on_error)
else:
self.retry = Retry(NoBackoff(), 0)
self.health_check_interval = health_check_interval
self.next_health_check: float = -1
self.encoder = encoder_class(encoding, encoding_errors, decode_responses)
self.redis_connect_func = redis_connect_func
self._reader: Optional[asyncio.StreamReader] = None
self._writer: Optional[asyncio.StreamWriter] = None
self._socket_read_size = socket_read_size
self._active_read_timeout = None
self._connect_callbacks: List[weakref.WeakMethod[ConnectCallbackT]] = []
self._buffer_cutoff = 6000
self._re_auth_token: Optional[TokenInterface] = None
self._should_reconnect = False
try:
p = int(protocol)
except TypeError:
p = DEFAULT_RESP_VERSION
except ValueError:
raise ConnectionError("protocol must be an integer")
else:
if p < 2 or p > 3:
raise ConnectionError("protocol must be either 2 or 3")
self.protocol = p
self.legacy_responses = legacy_responses
if parser_class != _AsyncHiredisParser:
# The Python parsers are protocol-specific; hiredis supports both.
if self.protocol == 3 and parser_class == _AsyncRESP2Parser:
parser_class = _AsyncRESP3Parser
elif self.protocol == 2 and parser_class == _AsyncRESP3Parser:
parser_class = _AsyncRESP2Parser
self.set_parser(parser_class)
# HIMPORT client-side state. `himport_registry` is the shared client-level
# registry (empty if unconfigured) and persists across reconnects.
self.himport_registry = himport_registry
self._reset_himport_state()
AsyncMaintNotificationsAbstractConnection.__init__(
self,
maint_notifications_config,
maint_notifications_pool_handler,
maintenance_state,
maintenance_notification_hash,
orig_host_address,
orig_socket_timeout,
orig_socket_connect_timeout,
oss_cluster_maint_notifications_handler,
self._parser,
)
def __del__(self, _warnings: Any = warnings):
# For some reason, the individual streams don't get properly garbage
# collected and therefore produce no resource warnings. We add one
# here, in the same style as those from the stdlib.
if getattr(self, "_writer", None):
_warnings.warn(
f"unclosed Connection {self!r}", ResourceWarning, source=self
)
try:
asyncio.get_running_loop()
self._close()
except RuntimeError:
# No actions been taken if pool already closed.
pass
def _close(self):
"""
Internal method to silently close the connection without waiting
"""
if self._writer:
self._writer.close()
self._writer = self._reader = None
def __repr__(self):
repr_args = ",".join((f"{k}={v}" for k, v in self.repr_pieces()))
return f"<{self.__class__.__module__}.{self.__class__.__name__}({repr_args})>"
@abstractmethod
def repr_pieces(self):
pass
@property
def is_connected(self):
return self._reader is not None and self._writer is not None
def register_connect_callback(self, callback):
"""
Register a callback to be called when the connection is established either
initially or reconnected. This allows listeners to issue commands that
are ephemeral to the connection, for example pub/sub subscription or
key tracking. The callback must be a _method_ and will be kept as
a weak reference.
"""
wm = weakref.WeakMethod(callback)
if wm not in self._connect_callbacks:
self._connect_callbacks.append(wm)
def deregister_connect_callback(self, callback):
"""
De-register a previously registered callback. It will no-longer receive
notifications on connection events. Calling this is not required when the
listener goes away, since the callbacks are kept as weak methods.
"""
try:
self._connect_callbacks.remove(weakref.WeakMethod(callback))
except ValueError:
pass
def set_parser(self, parser_class: Type[BaseParser]) -> None:
"""
Creates a new instance of parser_class with socket size:
_socket_read_size and assigns it to the parser for the connection
:param parser_class: The required parser class
"""
self._parser = parser_class(socket_read_size=self._socket_read_size)
def _get_parser(self) -> BaseParser:
return self._parser
def getpeername(self) -> str | None:
"""
Returns the peer name of the connection.
"""
writer = self._writer
if writer is None:
return None
peername = writer.get_extra_info("peername")
if isinstance(peername, tuple) and peername:
return str(peername[0])
return None
async def connect(self):
"""Connects to the Redis server if not already connected"""
# try once the socket connect with the handshake, retry the whole
# connect/handshake flow based on retry policy
await self.retry.call_with_retry(
lambda: self.connect_check_health(
check_health=True, retry_socket_connect=False
),
lambda error, failure_count: self.disconnect(
error=error, failure_count=failure_count
),
with_failure_count=True,
)
async def connect_check_health(
self, check_health: bool = True, retry_socket_connect: bool = True
):
if self.is_connected:
return
# Track actual retry attempts for error reporting
actual_retry_attempts = 0
def failure_callback(error, failure_count):
nonlocal actual_retry_attempts
actual_retry_attempts = failure_count
return self.disconnect(error=error, failure_count=failure_count)
try:
if retry_socket_connect:
await self.retry.call_with_retry(
lambda: self._connect(),
failure_callback,
with_failure_count=True,
)
else:
await self._connect()
except asyncio.CancelledError:
raise # in 3.7 and earlier, this is an Exception, not BaseException
except (socket.timeout, asyncio.TimeoutError):
e = TimeoutError("Timeout connecting to server")
await record_error_count(
server_address=getattr(self, "host", None),
server_port=getattr(self, "port", None),
network_peer_address=getattr(self, "host", None),
network_peer_port=getattr(self, "port", None),
error_type=e,
retry_attempts=actual_retry_attempts,
is_internal=False,
)
raise e
except OSError as e:
e = ConnectionError(self._error_message(e))
await record_error_count(
server_address=getattr(self, "host", None),
server_port=getattr(self, "port", None),
network_peer_address=getattr(self, "host", None),
network_peer_port=getattr(self, "port", None),
error_type=e,
retry_attempts=actual_retry_attempts,
is_internal=False,
)
raise e
except Exception as exc:
raise ConnectionError(exc) from exc
try:
if not self.redis_connect_func:
# Use the default on_connect function
await self.on_connect_check_health(check_health=check_health)
else:
# Use the passed function redis_connect_func
(
await self.redis_connect_func(self)
if asyncio.iscoroutinefunction(self.redis_connect_func)
else self.redis_connect_func(self)
)
except RedisError:
# clean up after any error in on_connect
await self.disconnect()
raise
# run any user callbacks. right now the only internal callback
# is for pubsub channel/pattern resubscription
# first, remove any dead weakrefs
self._connect_callbacks = [ref for ref in self._connect_callbacks if ref()]
for ref in self._connect_callbacks:
callback = ref()
task = callback(self)
if task and inspect.isawaitable(task):
await task
def mark_for_reconnect(self):
self._should_reconnect = True
def should_reconnect(self):
return self._should_reconnect
def reset_should_reconnect(self):
self._should_reconnect = False
@abstractmethod
async def _connect(self):
pass
@abstractmethod
def _host_error(self) -> str:
pass
def _error_message(self, exception: BaseException) -> str:
return format_error_message(self._host_error(), exception)
def get_protocol(self):
return self.protocol
def _reset_himport_state(self) -> None:
# A fresh server session has no prepared HIMPORT fieldsets, so the next
# himport_set must re-prepare on this connection. ``_himport_prepared`` maps
# fieldset name -> the version prepared on the server; ``_himport_reconciled
# _revision`` is the registry revision this connection last reconciled discards
# against. Both are reset on connect/disconnect since the session is gone.
self._himport_prepared: dict[str, int] = {}
self._himport_reconciled_revision: int = 0
async def on_connect(self) -> None:
"""Initialize the connection, authenticate and select a database"""
await self.on_connect_check_health(check_health=True)
async def on_connect_check_health(self, check_health: bool = True) -> None:
# A fresh socket is a new server session: no prepared HIMPORT fieldsets.
self._reset_himport_state()
self._parser.on_connect(self)
parser = self._parser
auth_args = None
# if credential provider or username and/or password are set, authenticate
if self.credential_provider or (self.username or self.password):
cred_provider = (
self.credential_provider
or UsernamePasswordCredentialProvider(self.username, self.password)
)
auth_args = await cred_provider.get_credentials_async()
# if resp version is specified and we have auth args,
# we need to send them via HELLO
if auth_args and check_protocol_version(self.protocol, 3):
if isinstance(self._parser, _AsyncRESP2Parser):
self.set_parser(_AsyncRESP3Parser)
# update cluster exception classes
self._parser.EXCEPTION_CLASSES = parser.EXCEPTION_CLASSES
self._parser.on_connect(self)
if len(auth_args) == 1:
auth_args = ["default", auth_args[0]]
# avoid checking health here -- PING will fail if we try
# to check the health prior to the AUTH
await self.send_command(
"HELLO", self.protocol, "AUTH", *auth_args, check_health=False
)
response = await self.read_response()
if response.get(b"proto") != int(self.protocol) and response.get(
"proto"
) != int(self.protocol):
raise ConnectionError("Invalid RESP version")
# avoid checking health here -- PING will fail if we try
# to check the health prior to the AUTH
elif auth_args:
await self.send_command("AUTH", *auth_args, check_health=False)
try:
auth_response = await self.read_response()
except AuthenticationWrongNumberOfArgsError:
# a username and password were specified but the Redis
# server seems to be < 6.0.0 which expects a single password
# arg. retry auth with just the password.
# https://github.com/andymccurdy/redis-py/issues/1274
await self.send_command("AUTH", auth_args[-1], check_health=False)
auth_response = await self.read_response()
if str_if_bytes(auth_response) != "OK":
raise AuthenticationError("Invalid Username or Password")
# if resp version is specified, switch to it
elif check_protocol_version(self.protocol, 3):
if isinstance(self._parser, _AsyncRESP2Parser):
self.set_parser(_AsyncRESP3Parser)
# update cluster exception classes
self._parser.EXCEPTION_CLASSES = parser.EXCEPTION_CLASSES
self._parser.on_connect(self)
await self.send_command("HELLO", self.protocol, check_health=check_health)
response = await self.read_response()
# if response.get(b"proto") != self.protocol and response.get(
# "proto"
# ) != self.protocol:
# raise ConnectionError("Invalid RESP version")
# Activate maintenance notifications for this connection
# if enabled in the configuration
# This is a no-op if maintenance notifications are not enabled
await self.activate_maint_notifications_handling_if_enabled(
check_health=check_health
)
# if a client_name is given, set it
if self.client_name:
await self.send_command(
"CLIENT",
"SETNAME",
self.client_name,
check_health=check_health,
)
if str_if_bytes(await self.read_response()) != "OK":
raise ConnectionError("Error setting client name")
# Set the library name and version from driver_info, pipeline for lower startup latency
lib_name_sent = False
lib_version_sent = False
if self.driver_info and self.driver_info.formatted_name:
await self.send_command(
"CLIENT",
"SETINFO",
"LIB-NAME",
self.driver_info.formatted_name,
check_health=check_health,
)
lib_name_sent = True
if self.driver_info and self.driver_info.lib_version:
await self.send_command(
"CLIENT",
"SETINFO",
"LIB-VER",
self.driver_info.lib_version,
check_health=check_health,
)
lib_version_sent = True
# if a database is specified, switch to it. Also pipeline this
if self.db:
await self.send_command("SELECT", self.db, check_health=check_health)
# read responses from pipeline
for _ in range(sum([lib_name_sent, lib_version_sent])):
try:
await self.read_response()
except ResponseError:
pass
if self.db:
if str_if_bytes(await self.read_response()) != "OK":
raise ConnectionError("Invalid Database")
async def disconnect(
self,
nowait: bool = False,
error: Optional[Exception] = None,
failure_count: Optional[int] = None,
health_check_failed: bool = False,
) -> None:
"""Disconnects from the Redis server"""
# The server session is gone, so any HIMPORT fieldsets prepared on this
# socket no longer exist; reset the tracking.
self._reset_himport_state()
# On Python 3.13+, asyncio.timeout() raises RuntimeError when called
# outside a running Task (e.g. during GC finalization or event-loop
# callbacks). In that context we fall back to a synchronous close.
# See https://github.com/redis/redis-py/issues/3856
if asyncio.current_task() is None:
self._parser.on_disconnect()
self.reset_should_reconnect()
self._close()
return
try:
async with async_timeout(self.socket_connect_timeout):
self._parser.on_disconnect()
# Reset the reconnect flag
self.reset_should_reconnect()
if not self.is_connected:
return
try:
self._writer.close() # type: ignore[union-attr]
# wait for close to finish, except when handling errors and
# forcefully disconnecting.
if not nowait:
await self._writer.wait_closed() # type: ignore[union-attr]
except OSError:
pass
finally:
self._reader = None
self._writer = None
except asyncio.TimeoutError:
raise TimeoutError(
f"Timed out closing connection after {self.socket_connect_timeout}"
) from None
if error:
if health_check_failed:
close_reason = CloseReason.HEALTHCHECK_FAILED
else:
close_reason = CloseReason.ERROR
if failure_count is not None and failure_count > self.retry.get_retries():
await record_error_count(
server_address=getattr(self, "host", None),
server_port=getattr(self, "port", None),
network_peer_address=getattr(self, "host", None),
network_peer_port=getattr(self, "port", None),
error_type=error,
retry_attempts=failure_count,
)
await record_connection_closed(
close_reason=close_reason,
error_type=error,
)
else:
await record_connection_closed(
close_reason=CloseReason.APPLICATION_CLOSE,
)
if self.maintenance_state == MaintenanceState.MAINTENANCE:
# MOVING state is owned by the pool-level TTL cleanup. Regular
# maintenance timeout relaxation can be restored when this
# connection closes, matching the sync lifecycle.
self.reset_tmp_settings(reset_relaxed_timeout=True)
self.maintenance_state = MaintenanceState.NONE
# reset the sets that keep track of received start maint
# notifications and skipped end maint notifications
self.reset_received_notifications()
async def _send_ping(self):
"""Send PING, expect PONG in return"""
await self.send_command("PING", check_health=False)
if str_if_bytes(await self.read_response()) != "PONG":
raise ConnectionError("Bad response from PING health check")
async def _ping_failed(self, error, failure_count):
"""Function to call when PING fails"""
await self.disconnect(
error=error, failure_count=failure_count, health_check_failed=True
)
async def check_health(self):
"""Check the health of the connection with a PING/PONG"""
if (
self.health_check_interval
and asyncio.get_running_loop().time() > self.next_health_check
):
await self.retry.call_with_retry(
self._send_ping, self._ping_failed, with_failure_count=True
)
async def _send_packed_command(self, command: Iterable[bytes]) -> None:
self._writer.writelines(command)
await self._writer.drain()
async def send_packed_command(
self, command: Union[bytes, str, Iterable[bytes]], check_health: bool = True
) -> None:
if not self.is_connected:
await self.connect_check_health(check_health=False)
if check_health:
await self.check_health()
try:
if isinstance(command, str):
command = command.encode()
if isinstance(command, bytes):
command = [command]
if self.socket_timeout:
await asyncio.wait_for(
self._send_packed_command(command), self.socket_timeout
)
else:
self._writer.writelines(command)
await self._writer.drain()
except asyncio.TimeoutError:
await self.disconnect(nowait=True)
raise TimeoutError("Timeout writing to socket") from None
except OSError as e:
await self.disconnect(nowait=True)
if len(e.args) == 1:
err_no, errmsg = "UNKNOWN", e.args[0]
else:
err_no = e.args[0]
errmsg = e.args[1]
raise ConnectionError(
f"Error {err_no} while writing to socket. {errmsg}."
) from e
except BaseException:
# BaseExceptions can be raised when a socket send operation is not
# finished, e.g. due to a timeout. Ideally, a caller could then re-try
# to send un-sent data. However, the send_packed_command() API
# does not support it so there is no point in keeping the connection open.
await self.disconnect(nowait=True)
raise
async def send_command(self, *args: Any, **kwargs: Any) -> None:
"""Pack and send a command to the Redis server"""
await self.send_packed_command(
self.pack_command(*args), check_health=kwargs.get("check_health", True)
)
@deprecated_function(
version="8.0.0", reason="Use can_read() instead", name="can_read_destructive"
)
async def can_read_destructive(self) -> bool:
"""Check the socket to see if there's data loaded in the buffer."""
try:
return await self._parser.can_read()
except OSError as e:
await self.disconnect(nowait=True)
host_error = self._host_error()
raise ConnectionError(f"Error while reading from {host_error}: {e.args}")
async def can_read(self) -> bool:
"""Check the socket to see if there's data loaded in the buffer."""
# TODO: Rename this API; it detects pending data or dirty/closed
# connection state, not only whether application data can be read.
try:
return await self._parser.can_read()
except OSError as e:
await self.disconnect(nowait=True)
host_error = self._host_error()
raise ConnectionError(f"Error while reading from {host_error}: {e.args}")
async def read_response(
self,
disable_decoding: bool = False,
timeout: float | None = None,
*,
disconnect_on_error: bool = True,
push_request: bool | None = False,
):
"""Read the response from a previously sent command.
``timeout`` semantics:
- ``None`` (default): fall back to ``self.socket_timeout``.
- ``math.inf``: block indefinitely with no timeout. Used by PubSub
blocking reads (``listen()`` / ``get_message(timeout=None)`` /
``parse_response(block=True)``) where the configured
``socket_timeout`` must not abort the read.
- ``float``: apply that timeout in seconds for this single read.
TODO(next-major): replace the ``math.inf`` opt-in with a SENTINEL
default for ``timeout``. After that change, ``timeout=None`` will
mean "no timeout, block until a response arrives" (matching the
long-standing PubSub docstring contract) and the SENTINEL default
will be the value that falls back to ``self.socket_timeout``.
That swap is a breaking change, so it must wait for a major
release. Until then, callers that need an indefinitely blocking
read pass ``math.inf`` explicitly.
"""
# TODO(next-major): drop the math.inf branch. Use SENTINEL as the
# default for ``timeout`` and treat ``timeout is None`` as the
# "no timeout" signal (matching the PubSub docstring contract).
# Match only positive infinity here. ``-math.inf`` is not a valid
# "block forever" signal and historically behaved as an already-
# expired timeout; preserve that.
if timeout == math.inf:
read_timeout = None
else:
read_timeout = timeout if timeout is not None else self.socket_timeout
host_error = self._host_error()
try:
if read_timeout is not None:
timeout_context = async_timeout(read_timeout)
if timeout is None:
async with timeout_context as active_timeout:
self._active_read_timeout = active_timeout
try:
response = await self._read_response_from_parser(
disable_decoding=disable_decoding,
push_request=push_request,
)
finally:
self._active_read_timeout = None
else:
async with timeout_context:
response = await self._read_response_from_parser(
disable_decoding=disable_decoding,
push_request=push_request,
)
else:
response = await self._read_response_from_parser(
disable_decoding=disable_decoding,
push_request=push_request,
)
except asyncio.TimeoutError:
if timeout is not None:
# user requested timeout, return None. Operation can be retried
return None
# it was a self.socket_timeout error.
if disconnect_on_error:
await self.disconnect(nowait=True)
raise TimeoutError(f"Timeout reading from {host_error}")
except OSError as e:
if disconnect_on_error:
await self.disconnect(nowait=True)
raise ConnectionError(f"Error while reading from {host_error} : {e.args}")
except BaseException:
# Also by default close in case of BaseException. A lot of code
# relies on this behaviour when doing Command/Response pairs.
# See #1128.
if disconnect_on_error:
await self.disconnect(nowait=True)
raise
if self.health_check_interval:
next_time = asyncio.get_running_loop().time() + self.health_check_interval
self.next_health_check = next_time
if isinstance(response, ResponseError):
raise response from None
return response
async def _read_response_from_parser(
self, disable_decoding: bool = False, push_request: bool | None = False
):
if check_protocol_version(self.protocol, 3):
return await self._parser.read_response(
disable_decoding=disable_decoding, push_request=push_request
)
return await self._parser.read_response(disable_decoding=disable_decoding)
def pack_command(self, *args: EncodableT) -> List[bytes]:
"""Pack a series of arguments into the Redis protocol"""
output = []
# the client might have included 1 or more literal arguments in
# the command name, e.g., 'CONFIG GET'. The Redis server expects these
# arguments to be sent separately, so split the first argument
# manually. These arguments should be bytestrings so that they are
# not encoded.
assert not isinstance(args[0], float)
if isinstance(args[0], str):
args = tuple(args[0].encode().split()) + args[1:]
elif b" " in args[0]:
args = tuple(args[0].split()) + args[1:]
buff = SYM_EMPTY.join((SYM_STAR, str(len(args)).encode(), SYM_CRLF))
buffer_cutoff = self._buffer_cutoff
for arg in map(self.encoder.encode, args):
# to avoid large string mallocs, chunk the command into the
# output list if we're sending large values or memoryviews
arg_length = len(arg)
if (
len(buff) > buffer_cutoff
or arg_length > buffer_cutoff
or isinstance(arg, memoryview)
):
buff = SYM_EMPTY.join(
(buff, SYM_DOLLAR, str(arg_length).encode(), SYM_CRLF)
)
output.append(buff)
output.append(arg)
buff = SYM_CRLF
else:
buff = SYM_EMPTY.join(
(
buff,
SYM_DOLLAR,
str(arg_length).encode(),
SYM_CRLF,
arg,
SYM_CRLF,
)
)
output.append(buff)
return output
def pack_commands(self, commands: Iterable[Iterable[EncodableT]]) -> List[bytes]:
"""Pack multiple commands into the Redis protocol"""
output: List[bytes] = []
pieces: List[bytes] = []
buffer_length = 0
buffer_cutoff = self._buffer_cutoff
for cmd in commands:
for chunk in self.pack_command(*cmd):
chunklen = len(chunk)
if (
buffer_length > buffer_cutoff
or chunklen > buffer_cutoff
or isinstance(chunk, memoryview)
):
if pieces:
output.append(SYM_EMPTY.join(pieces))
buffer_length = 0
pieces = []
if chunklen > buffer_cutoff or isinstance(chunk, memoryview):
output.append(chunk)
else:
pieces.append(chunk)
buffer_length += chunklen
if pieces:
output.append(SYM_EMPTY.join(pieces))
return output
def _socket_is_empty(self):
"""Check if the socket is empty"""
return len(self._reader._buffer) == 0
async def process_invalidation_messages(self):
while not self._socket_is_empty():
await self.read_response(push_request=True)
def set_re_auth_token(self, token: TokenInterface):
self._re_auth_token = token
async def re_auth(self):
if self._re_auth_token is not None:
await self.send_command(
"AUTH",
self._re_auth_token.try_get("oid"),
self._re_auth_token.get_value(),
)
await self.read_response()
self._re_auth_token = None
[docs]class Connection(AbstractConnection):
"Manages TCP communication to and from a Redis server"
def __init__(
self,
*,
host: str = "localhost",
port: str | int = 6379,
socket_keepalive: bool = True,
socket_keepalive_options: Mapping[int, int | bytes] | object | None = SENTINEL,
socket_type: int = 0,
**kwargs,
):
"""
Initialize a TCP connection.
Parameters
----------
socket_keepalive : bool
If `True`, TCP keepalive is enabled for TCP socket connections.
socket_keepalive_options : Mapping[int, int | bytes] | object | None
Mapping of TCP keepalive socket option constants to values, for
example `{socket.TCP_KEEPIDLE: 30}`. If left unspecified, redis-py
uses TCP keepalive defaults when `socket_keepalive` is enabled:
idle 30 seconds, interval 5 seconds, and 3 probes. Platform-specific
options that are not available are skipped. Pass `None` or `{}` to
avoid setting additional TCP keepalive options.
"""
self.host = host
self.port = int(port)
self.socket_keepalive = socket_keepalive
if socket_keepalive_options is SENTINEL:
socket_keepalive_options = get_default_socket_keepalive_options()
self.socket_keepalive_options = socket_keepalive_options or {}
self.socket_type = socket_type
super().__init__(**kwargs)
def repr_pieces(self):
pieces = [("host", self.host), ("port", self.port), ("db", self.db)]
if self.client_name:
pieces.append(("client_name", self.client_name))
return pieces
def _connection_arguments(self) -> Mapping:
return {"host": self.host, "port": self.port}
async def _connect(self):
"""Create a TCP socket connection"""
async with async_timeout(self.socket_connect_timeout):
reader, writer = await asyncio.open_connection(
**self._connection_arguments()
)
self._reader = reader
self._writer = writer
sock = writer.transport.get_extra_info("socket")
if sock:
sock.setsockopt(socket.IPPROTO_TCP, socket.TCP_NODELAY, 1)
try:
# TCP_KEEPALIVE
if self.socket_keepalive:
sock.setsockopt(socket.SOL_SOCKET, socket.SO_KEEPALIVE, 1)
for k, v in self.socket_keepalive_options.items():
sock.setsockopt(socket.SOL_TCP, k, v)
except (OSError, TypeError):
# `socket_keepalive_options` might contain invalid options
# causing an error. Do not leave the connection open.
writer.close()
raise
def _host_error(self) -> str:
return f"{self.host}:{self.port}"
class SSLConnection(Connection):
"""Manages SSL connections to and from the Redis server(s).
This class extends the Connection class, adding SSL functionality, and making
use of ssl.SSLContext (https://docs.python.org/3/library/ssl.html#ssl.SSLContext)
"""
def __init__(
self,
ssl_keyfile: Optional[str] = None,
ssl_certfile: Optional[str] = None,
ssl_cert_reqs: Union[str, ssl.VerifyMode] = "required",
ssl_include_verify_flags: Optional[List["ssl.VerifyFlags"]] = None,
ssl_exclude_verify_flags: Optional[List["ssl.VerifyFlags"]] = None,
ssl_ca_certs: Optional[str] = None,
ssl_ca_data: Optional[str] = None,
ssl_ca_path: Optional[str] = None,
ssl_check_hostname: bool = True,
ssl_min_version: Optional[TLSVersion] = None,
ssl_ciphers: Optional[str] = None,
ssl_password: Optional[str] = None,
**kwargs,
):
if not SSL_AVAILABLE:
raise RedisError("Python wasn't built with SSL support")
self.ssl_context: RedisSSLContext = RedisSSLContext(
keyfile=ssl_keyfile,
certfile=ssl_certfile,
cert_reqs=ssl_cert_reqs,
include_verify_flags=ssl_include_verify_flags,
exclude_verify_flags=ssl_exclude_verify_flags,
ca_certs=ssl_ca_certs,
ca_data=ssl_ca_data,
ca_path=ssl_ca_path,
check_hostname=ssl_check_hostname,
min_version=ssl_min_version,
ciphers=ssl_ciphers,
password=ssl_password,
)
super().__init__(**kwargs)
def _connection_arguments(self) -> Mapping:
kwargs = super()._connection_arguments()
kwargs["ssl"] = self.ssl_context.get()
return kwargs
@property
def keyfile(self):
return self.ssl_context.keyfile
@property
def certfile(self):
return self.ssl_context.certfile
@property
def cert_reqs(self):
return self.ssl_context.cert_reqs
@property
def include_verify_flags(self):
return self.ssl_context.include_verify_flags
@property
def exclude_verify_flags(self):
return self.ssl_context.exclude_verify_flags
@property
def ca_certs(self):
return self.ssl_context.ca_certs
@property
def ca_data(self):
return self.ssl_context.ca_data
@property
def check_hostname(self):
return self.ssl_context.check_hostname
@property
def min_version(self):
return self.ssl_context.min_version
class RedisSSLContext:
__slots__ = (
"keyfile",
"certfile",
"cert_reqs",
"include_verify_flags",
"exclude_verify_flags",
"ca_certs",
"ca_data",
"ca_path",
"context",
"check_hostname",
"min_version",
"ciphers",
"password",
)
def __init__(
self,
keyfile: Optional[str] = None,
certfile: Optional[str] = None,
cert_reqs: Optional[Union[str, ssl.VerifyMode]] = None,
include_verify_flags: Optional[List["ssl.VerifyFlags"]] = None,
exclude_verify_flags: Optional[List["ssl.VerifyFlags"]] = None,
ca_certs: Optional[str] = None,
ca_data: Optional[str] = None,
ca_path: Optional[str] = None,
check_hostname: bool = False,
min_version: Optional[TLSVersion] = None,
ciphers: Optional[str] = None,
password: Optional[str] = None,
):
if not SSL_AVAILABLE:
raise RedisError("Python wasn't built with SSL support")
self.keyfile = keyfile
self.certfile = certfile
if cert_reqs is None:
cert_reqs = ssl.CERT_NONE
elif isinstance(cert_reqs, str):
CERT_REQS = { # noqa: N806
"none": ssl.CERT_NONE,
"optional": ssl.CERT_OPTIONAL,
"required": ssl.CERT_REQUIRED,
}
if cert_reqs not in CERT_REQS:
raise RedisError(
f"Invalid SSL Certificate Requirements Flag: {cert_reqs}"
)
cert_reqs = CERT_REQS[cert_reqs]
self.cert_reqs = cert_reqs
self.include_verify_flags = include_verify_flags
self.exclude_verify_flags = exclude_verify_flags
self.ca_certs = ca_certs
self.ca_data = ca_data
self.ca_path = ca_path
self.check_hostname = (
check_hostname if self.cert_reqs != ssl.CERT_NONE else False
)
self.min_version = min_version
self.ciphers = ciphers
self.password = password
self.context: Optional[SSLContext] = None
def get(self) -> SSLContext:
if not self.context:
context = ssl.create_default_context()
context.check_hostname = self.check_hostname
context.verify_mode = self.cert_reqs
if self.include_verify_flags:
for flag in self.include_verify_flags:
context.verify_flags |= flag
if self.exclude_verify_flags:
for flag in self.exclude_verify_flags:
context.verify_flags &= ~flag
if self.certfile or self.keyfile:
context.load_cert_chain(
certfile=self.certfile,
keyfile=self.keyfile,
password=self.password,
)
if self.ca_certs or self.ca_data or self.ca_path:
context.load_verify_locations(
cafile=self.ca_certs, capath=self.ca_path, cadata=self.ca_data
)
if self.min_version is not None:
context.minimum_version = self.min_version
if self.ciphers is not None:
context.set_ciphers(self.ciphers)
self.context = context
return self.context
class UnixDomainSocketConnection(AbstractConnection):
"Manages UDS communication to and from a Redis server"
def __init__(self, *, path: str = "", **kwargs):
self.path = path
super().__init__(**kwargs)
def repr_pieces(self) -> Iterable[Tuple[str, Union[str, int]]]:
pieces = [("path", self.path), ("db", self.db)]
if self.client_name:
pieces.append(("client_name", self.client_name))
return pieces
async def _connect(self):
async with async_timeout(self.socket_connect_timeout):
reader, writer = await asyncio.open_unix_connection(path=self.path)
self._reader = reader
self._writer = writer
await self.on_connect()
def _host_error(self) -> str:
return self.path
FALSE_STRINGS = ("0", "F", "FALSE", "N", "NO")
def to_bool(value) -> Optional[bool]:
if value is None or value == "":
return None
if isinstance(value, str) and value.upper() in FALSE_STRINGS:
return False
return bool(value)
def parse_ssl_verify_flags(value):
# flags are passed in as a string representation of a list,
# e.g. VERIFY_X509_STRICT, VERIFY_X509_PARTIAL_CHAIN
verify_flags_str = value.replace("[", "").replace("]", "")
verify_flags = []
for flag in verify_flags_str.split(","):
flag = flag.strip()
if not hasattr(VerifyFlags, flag):
raise ValueError(f"Invalid ssl verify flag: {flag}")
verify_flags.append(getattr(VerifyFlags, flag))
return verify_flags
URL_QUERY_ARGUMENT_PARSERS: Mapping[str, Callable[..., object]] = MappingProxyType(
{
"db": int,
"socket_timeout": float,
"socket_connect_timeout": float,
"socket_read_size": int,
"socket_keepalive": to_bool,
"retry_on_timeout": to_bool,
"max_connections": int,
"health_check_interval": int,
"ssl_check_hostname": to_bool,
"ssl_include_verify_flags": parse_ssl_verify_flags,
"ssl_exclude_verify_flags": parse_ssl_verify_flags,
"ssl_min_version": int,
"timeout": float,
"protocol": int,
"legacy_responses": to_bool,
}
)
class ConnectKwargs(TypedDict, total=False):
username: str
password: str
connection_class: Type[AbstractConnection]
host: str
port: int
db: int
path: str
def parse_url(url: str) -> ConnectKwargs:
parsed: ParseResult = urlparse(url)
kwargs: ConnectKwargs = {}
for name, value_list in parse_qs(parsed.query).items():
if value_list and len(value_list) > 0:
# parse_qs() already percent-decodes query values, so use the value
# as-is; unquoting again here would double-decode (e.g. "%2520" ->
# "%20" -> " "). See issue #4208.
value = value_list[0]
parser = URL_QUERY_ARGUMENT_PARSERS.get(name)
if parser:
try:
kwargs[name] = parser(value)
except (TypeError, ValueError):
raise ValueError(f"Invalid value for '{name}' in connection URL.")
else:
kwargs[name] = value
if parsed.username:
kwargs["username"] = unquote(parsed.username)
if parsed.password:
kwargs["password"] = unquote(parsed.password)
# We only support redis://, rediss:// and unix:// schemes.
if parsed.scheme == "unix":
if parsed.path:
kwargs["path"] = unquote(parsed.path)
kwargs["connection_class"] = UnixDomainSocketConnection
elif parsed.scheme in ("redis", "rediss"):
if parsed.hostname:
kwargs["host"] = unquote(parsed.hostname)
if parsed.port:
kwargs["port"] = int(parsed.port)
# If there's a path argument, use it as the db argument if a
# querystring value wasn't specified
if parsed.path and "db" not in kwargs:
try:
kwargs["db"] = int(unquote(parsed.path).replace("/", ""))
except (AttributeError, ValueError):
pass
if parsed.scheme == "rediss":
kwargs["connection_class"] = SSLConnection
else:
valid_schemes = "redis://, rediss://, unix://"
raise ValueError(
f"Redis URL must specify one of the following schemes ({valid_schemes})"
)
return kwargs
_CP = TypeVar("_CP", bound="ConnectionPool")
class ConnectionPoolInterface(ABC):
@abstractmethod
def get_protocol(self):
pass
@abstractmethod
def reset(self) -> None:
pass
@abstractmethod
@deprecated_args(
args_to_warn=["*"],
reason="Use get_connection() without args instead",
version="5.3.0",
)
async def get_connection(
self, command_name: Optional[str] = None, *keys: Any, **options: Any
) -> "AbstractConnection":
pass
@abstractmethod
def get_encoder(self) -> "Encoder":
pass
@abstractmethod
async def release(self, connection: "AbstractConnection") -> None:
pass
@abstractmethod
async def disconnect(self, inuse_connections: bool = True) -> None:
pass
@abstractmethod
async def aclose(self) -> None:
pass
@abstractmethod
def set_retry(self, retry: "Retry") -> None:
pass
@abstractmethod
async def re_auth_callback(self, token: TokenInterface) -> None:
pass
@abstractmethod
def get_connection_count(self) -> List[Tuple[int, dict]]:
"""
Returns a connection count (both idle and in use).
"""
pass
class AsyncMaintNotificationsAbstractConnectionPool:
"""
Internal mixin for async maintenance notification pool wiring.
The handler owns notification policy.
This mixin owns pool state mutation because `_available_connections`,
`_in_use_connections`, `connection_kwargs`, and the non-reentrant `asyncio.Lock`
all live on the pool.
"""
def __init__(
self,
maint_notifications_config: MaintNotificationsConfig | None = None,
oss_cluster_maint_notifications_handler: (
"AsyncOSSMaintNotificationsHandler | None"
) = None,
**kwargs: Any,
) -> None:
protocol = kwargs.get("protocol")
is_protocol_supported = check_protocol_version(protocol, 3)
is_connection_supported = self._maintenance_notifications_supported()
if (
maint_notifications_config is None
and is_protocol_supported
and is_connection_supported
):
maint_notifications_config = MaintNotificationsConfig()
if maint_notifications_config and maint_notifications_config.enabled:
if not is_connection_supported:
if maint_notifications_config.enabled is True:
# Unix sockets do not have a host endpoint for CLIENT
# MAINT_NOTIFICATIONS to describe.
if "path" in self.connection_kwargs:
raise RedisError(
"Maintenance notifications are not supported for "
"Unix domain socket connections"
)
# Custom connection classes must inherit the async maintenance
# mixin so handlers can update connection state safely.
if not self._maintenance_notifications_connection_class_supported():
connection_class = getattr(self, "connection_class", None)
connection_class_name = getattr(
connection_class, "__name__", connection_class
)
raise RedisError(
"Maintenance notifications are not supported for "
f"connection class {connection_class_name}"
)
# TCP-like connections still need a host to identify the
# endpoint that can move during maintenance.
raise RedisError(
"Maintenance notifications are not supported for connections "
"without a host"
)
self._maint_notifications_pool_handler = None
self._oss_cluster_maint_notifications_handler = None
return
if not is_protocol_supported:
raise RedisError(
"Maintenance notifications handlers on connection are only supported with RESP version 3"
)
if oss_cluster_maint_notifications_handler:
self._oss_cluster_maint_notifications_handler = (
oss_cluster_maint_notifications_handler
)
self._update_connection_kwargs_for_maint_notifications(
oss_cluster_maint_notifications_handler=self._oss_cluster_maint_notifications_handler
)
self._maint_notifications_pool_handler = None
else:
self._oss_cluster_maint_notifications_handler = None
self._maint_notifications_pool_handler = (
AsyncMaintNotificationsPoolHandler(self, maint_notifications_config)
)
self._update_connection_kwargs_for_maint_notifications(
maint_notifications_pool_handler=self._maint_notifications_pool_handler
)
else:
self._maint_notifications_pool_handler = None
self._oss_cluster_maint_notifications_handler = None
async def _on_close(self) -> None:
"""Hook invoked from the pool's ``aclose()`` before the pool is shut down."""
if self._maint_notifications_pool_handler is not None:
await self._maint_notifications_pool_handler.cancel_scheduled_tasks()
@property
@abstractmethod
def connection_kwargs(self) -> dict[str, Any]:
pass
@connection_kwargs.setter
@abstractmethod
def connection_kwargs(self, value: dict[str, Any]) -> None:
pass
@abstractmethod
def _get_pool_lock(self) -> asyncio.Lock:
pass
@abstractmethod
def _get_free_connections(self) -> Iterable["AbstractConnection"]:
pass
@abstractmethod
def _get_in_use_connections(self) -> Iterable["AbstractConnection"]:
pass
def _maintenance_notifications_supported(self) -> bool:
if "path" in self.connection_kwargs:
return False
if not self._maintenance_notifications_connection_class_supported():
return False
return bool(self.connection_kwargs.get("host"))
def _maintenance_notifications_connection_class_supported(self) -> bool:
connection_class = getattr(self, "connection_class", None)
if connection_class is None:
return False
try:
return issubclass(
connection_class, AsyncMaintNotificationsAbstractConnection
)
except TypeError:
return False
def maint_notifications_enabled(self):
"""
Returns:
True if the maintenance notifications are enabled, False otherwise.
The maintenance notifications config is stored in the pool handler.
If the pool handler is not set, the maintenance notifications are not enabled.
"""
if self._oss_cluster_maint_notifications_handler:
maint_notifications_config = (
self._oss_cluster_maint_notifications_handler.config
)
else:
maint_notifications_config = (
self._maint_notifications_pool_handler.config
if self._maint_notifications_pool_handler
else None
)
return maint_notifications_config and maint_notifications_config.enabled
async def update_maint_notifications_config(
self,
maint_notifications_config: MaintNotificationsConfig,
oss_cluster_maint_notifications_handler: (
AsyncOSSMaintNotificationsHandler | None
) = None,
) -> None:
"""
Updates the maintenance notifications configuration.
This method should be called only if the pool was created
without enabling the maintenance notifications and
in a later point in time maintenance notifications
are requested to be enabled.
"""
if (
self.maint_notifications_enabled()
and not maint_notifications_config.enabled
):
raise ValueError(
"Cannot disable maintenance notifications after enabling them"
)
if oss_cluster_maint_notifications_handler:
self._oss_cluster_maint_notifications_handler = (
oss_cluster_maint_notifications_handler
)
# OSS cluster mode and pool-handler mode are mutually exclusive
# (see __init__). A pool created with the default RESP3 "auto"
# config wires a pool handler before this method runs; clear it so
# new and existing connections are not configured with both handlers.
self._maint_notifications_pool_handler = None
else:
if (
maint_notifications_config.enabled
and not self._maintenance_notifications_supported()
):
if maint_notifications_config.enabled is True:
# Unix sockets do not have a host endpoint for CLIENT
# MAINT_NOTIFICATIONS to describe.
if "path" in self.connection_kwargs:
raise RedisError(
"Maintenance notifications are not supported for "
"Unix domain socket connections"
)
# Custom connection classes must inherit the async maintenance
# mixin so handlers can update connection state safely.
if not self._maintenance_notifications_connection_class_supported():
connection_class = getattr(self, "connection_class", None)
connection_class_name = getattr(
connection_class, "__name__", connection_class
)
raise RedisError(
"Maintenance notifications are not supported for "
f"connection class {connection_class_name}"
)
# TCP-like connections still need a host to identify the
# endpoint that can move during maintenance.
raise RedisError(
"Maintenance notifications are not supported for connections "
"without a host"
)
self._maint_notifications_pool_handler = None
return
if self._oss_cluster_maint_notifications_handler:
# Pool already in OSS cluster mode; update the OSS handler config
# instead of creating a mutually-exclusive pool handler (which
# would be silently ignored because the OSS handler wins priority
# in both update helpers below).
self._oss_cluster_maint_notifications_handler.config = (
maint_notifications_config
)
elif not self._maint_notifications_pool_handler:
self._maint_notifications_pool_handler = (
AsyncMaintNotificationsPoolHandler(self, maint_notifications_config)
)
else:
self._maint_notifications_pool_handler.config = (
maint_notifications_config
)
self._update_connection_kwargs_for_maint_notifications(
maint_notifications_pool_handler=self._maint_notifications_pool_handler,
oss_cluster_maint_notifications_handler=self._oss_cluster_maint_notifications_handler,
)
await self._update_maint_notifications_configs_for_connections(
maint_notifications_pool_handler=self._maint_notifications_pool_handler,
oss_cluster_maint_notifications_handler=self._oss_cluster_maint_notifications_handler,
)
def _update_connection_kwargs_for_maint_notifications(
self,
maint_notifications_pool_handler: (
AsyncMaintNotificationsPoolHandler | None
) = None,
oss_cluster_maint_notifications_handler: (
AsyncOSSMaintNotificationsHandler | None
) = None,
) -> None:
"""
Update the connection kwargs for all future connections.
"""
if not self.maint_notifications_enabled():
return
if maint_notifications_pool_handler:
self.connection_kwargs.update(
{
"maint_notifications_pool_handler": maint_notifications_pool_handler,
"maint_notifications_config": maint_notifications_pool_handler.config,
}
)
if oss_cluster_maint_notifications_handler:
self.connection_kwargs.update(
{
"oss_cluster_maint_notifications_handler": oss_cluster_maint_notifications_handler,
"maint_notifications_config": oss_cluster_maint_notifications_handler.config,
}
)
# OSS cluster mode and pool-handler mode are mutually exclusive.
# Drop any pool handler a default (RESP3 "auto") pool creation may
# have wired so future connections are not configured with both.
self.connection_kwargs.pop("maint_notifications_pool_handler", None)
# Store original connection parameters for maintenance notifications.
if self.connection_kwargs.get("orig_host_address", None) is None:
# If orig_host_address is None it means we haven't
# configured the original values yet
self.connection_kwargs.update(
{
"orig_host_address": self.connection_kwargs.get("host"),
"orig_socket_timeout": self.connection_kwargs.get(
"socket_timeout", DEFAULT_SOCKET_TIMEOUT
),
"orig_socket_connect_timeout": self.connection_kwargs.get(
"socket_connect_timeout", DEFAULT_SOCKET_CONNECT_TIMEOUT
),
}
)
async def _update_maint_notifications_configs_for_connections(
self,
maint_notifications_pool_handler: (
AsyncMaintNotificationsPoolHandler | None
) = None,
oss_cluster_maint_notifications_handler: (
AsyncOSSMaintNotificationsHandler | None
) = None,
) -> None:
"""Update the maintenance notifications config for all connections in the pool."""
async with self._get_pool_lock():
for conn in list(self._get_free_connections()):
if oss_cluster_maint_notifications_handler:
conn.set_maint_notifications_cluster_handler_for_connection(
oss_cluster_maint_notifications_handler
)
conn.maint_notifications_config = (
oss_cluster_maint_notifications_handler.config
)
elif maint_notifications_pool_handler:
conn.set_maint_notifications_pool_handler_for_connection(
maint_notifications_pool_handler
)
conn.maint_notifications_config = (
maint_notifications_pool_handler.config
)
else:
raise ValueError(
"Either maint_notifications_pool_handler or "
"oss_cluster_maint_notifications_handler must be set"
)
await conn.disconnect()
for conn in list(self._get_in_use_connections()):
if oss_cluster_maint_notifications_handler:
# Use set_maint_notifications_cluster_handler_for_connection
# (not _configure_maintenance_notifications) so the parser is
# obtained from the connection itself. _configure_* requires a
# parser argument and would raise here; it would also reset the
# connection's orig_* settings, which is wrong for an in-use
# (active) connection. This mirrors the idle-connection branch
# above and the pool-handler branches.
conn.set_maint_notifications_cluster_handler_for_connection(
oss_cluster_maint_notifications_handler
)
conn.maint_notifications_config = (
oss_cluster_maint_notifications_handler.config
)
elif maint_notifications_pool_handler:
conn.set_maint_notifications_pool_handler_for_connection(
maint_notifications_pool_handler
)
conn.maint_notifications_config = (
maint_notifications_pool_handler.config
)
else:
raise ValueError(
"Either maint_notifications_pool_handler or "
"oss_cluster_maint_notifications_handler must be set"
)
conn.mark_for_reconnect()
def _should_update_connection(
self,
conn: "AbstractConnection",
matching_pattern: str = "connected_address",
matching_address: str | None = None,
matching_notification_hash: int | None = None,
) -> bool:
"""
Check if the connection should be updated based on the matching criteria.
"""
if matching_pattern == "connected_address":
if matching_address and conn.getpeername() != matching_address:
return False
elif matching_pattern == "configured_address":
if matching_address and conn.host != matching_address:
return False
elif matching_pattern == "notification_hash":
if (
matching_notification_hash is not None
and conn.maintenance_notification_hash != matching_notification_hash
):
return False
return True
def update_connection_settings(
self,
conn: "AsyncMaintNotificationsAbstractConnection",
state: MaintenanceState | None = None,
maintenance_notification_hash: int | None = None,
host_address: str | None = None,
relaxed_timeout: float | None = None,
update_notification_hash: bool = False,
reset_host_address: bool = False,
reset_relaxed_timeout: bool = False,
) -> None:
"""
Update the settings for a single connection.
"""
if state:
conn.maintenance_state = state
if update_notification_hash:
# update the notification hash only if requested
conn.maintenance_notification_hash = maintenance_notification_hash
if host_address is not None:
conn.set_tmp_settings(tmp_host_address=host_address)
if relaxed_timeout is not None:
conn.set_tmp_settings(tmp_relaxed_timeout=relaxed_timeout)
if reset_relaxed_timeout or reset_host_address:
conn.reset_tmp_settings(
reset_host_address=reset_host_address,
reset_relaxed_timeout=reset_relaxed_timeout,
)
conn.update_current_socket_timeout(relaxed_timeout)
async def update_connections_settings(
self,
state: MaintenanceState | None = None,
maintenance_notification_hash: int | None = None,
host_address: str | None = None,
relaxed_timeout: float | None = None,
matching_address: str | None = None,
matching_notification_hash: int | None = None,
matching_pattern: Literal[
"connected_address", "configured_address", "notification_hash"
] = "connected_address",
update_notification_hash: bool = False,
reset_host_address: bool = False,
reset_relaxed_timeout: bool = False,
include_free_connections: bool = True,
) -> None:
"""
Update the settings for all matching connections in the pool.
This method does not create new connections.
This method does not affect the connection kwargs.
:param state: The maintenance state to set for the connection.
:param maintenance_notification_hash: The hash of the maintenance notification
to set for the connection.
:param host_address: The host address to set for the connection.
:param relaxed_timeout: The relaxed timeout to set for the connection.
:param matching_address: The address to match for the connection.
:param matching_notification_hash: The notification hash to match for the connection.
:param matching_pattern: The pattern to match for the connection.
:param update_notification_hash: Whether to update the notification hash for the connection.
:param reset_host_address: Whether to reset the host address to the original address.
:param reset_relaxed_timeout: Whether to reset the relaxed timeout to the original timeout.
:param include_free_connections: Whether to include free/available connections.
"""
async with self._get_pool_lock():
self._update_connections_settings_without_locking(
state=state,
maintenance_notification_hash=maintenance_notification_hash,
host_address=host_address,
relaxed_timeout=relaxed_timeout,
matching_address=matching_address,
matching_notification_hash=matching_notification_hash,
matching_pattern=matching_pattern,
update_notification_hash=update_notification_hash,
reset_host_address=reset_host_address,
reset_relaxed_timeout=reset_relaxed_timeout,
include_free_connections=include_free_connections,
)
def _update_connections_settings_without_locking(
self,
state: MaintenanceState | None = None,
maintenance_notification_hash: int | None = None,
host_address: str | None = None,
relaxed_timeout: float | None = None,
matching_address: str | None = None,
matching_notification_hash: int | None = None,
matching_pattern: Literal[
"connected_address", "configured_address", "notification_hash"
] = "connected_address",
update_notification_hash: bool = False,
reset_host_address: bool = False,
reset_relaxed_timeout: bool = False,
include_free_connections: bool = True,
) -> None:
"""
Update matching connections while the caller already holds the pool lock.
This helper intentionally does not acquire the pool lock so callers can
compose several pool mutations inside one critical section without
deadlocking the non-reentrant `asyncio.Lock`.
"""
for conn in self._get_in_use_connections():
if self._should_update_connection(
conn,
matching_pattern,
matching_address,
matching_notification_hash,
):
self.update_connection_settings(
conn,
state=state,
maintenance_notification_hash=maintenance_notification_hash,
host_address=host_address,
relaxed_timeout=relaxed_timeout,
update_notification_hash=update_notification_hash,
reset_host_address=reset_host_address,
reset_relaxed_timeout=reset_relaxed_timeout,
)
if include_free_connections:
for conn in self._get_free_connections():
if self._should_update_connection(
conn,
matching_pattern,
matching_address,
matching_notification_hash,
):
self.update_connection_settings(
conn,
state=state,
maintenance_notification_hash=maintenance_notification_hash,
host_address=host_address,
relaxed_timeout=relaxed_timeout,
update_notification_hash=update_notification_hash,
reset_host_address=reset_host_address,
reset_relaxed_timeout=reset_relaxed_timeout,
)
def update_connection_kwargs(self, **kwargs: Any) -> None:
"""
Update the connection kwargs for all future connections.
This method updates the connection kwargs for all future connections created by the pool.
Existing connections are not affected.
"""
self.connection_kwargs.update(kwargs)
async def apply_moving_notification(
self,
notification: NodeMovingNotification,
config: MaintNotificationsConfig,
moving_address_src: str | None,
run_proactive_reconnect: bool = False,
) -> None:
"""
Apply the pool state transition for a MOVING notification atomically.
Async pools use a non-reentrant `asyncio.Lock`, so the handler cannot
safely compose several separately locked calls. Existing connection
updates, optional proactive reconnect, and future `connection_kwargs`
changes must happen under one pool-owned lock; otherwise a connection
can move between active/free lists and escape handling.
"""
async with self._get_pool_lock():
# Opt BlockingConnectionPool into serializing its get/release
# with this critical section. Other pools do not define
# set_in_maintenance and this is a no-op for them.
self._set_in_maintenance(True)
try:
self._update_connections_settings_without_locking(
state=MaintenanceState.MOVING,
maintenance_notification_hash=hash(notification),
relaxed_timeout=config.relaxed_timeout,
host_address=notification.new_node_host,
matching_address=moving_address_src,
matching_pattern="connected_address",
update_notification_hash=True,
include_free_connections=True,
)
if run_proactive_reconnect:
await self._run_proactive_reconnect_without_locking(
moving_address_src
)
self.update_connection_kwargs(
**_build_moving_connection_kwargs(notification, config)
)
finally:
self._set_in_maintenance(False)
async def run_proactive_reconnect(
self,
moving_address_src: str | None = None,
) -> None:
"""
Mark active connections and disconnect free connections atomically.
This operation is pool-owned because the active/free lists can change
while tasks acquire or release connections. Keeping the mark/disconnect
pass under one lock avoids a connection moving between lists between
separately locked calls.
"""
async with self._get_pool_lock():
await self._run_proactive_reconnect_without_locking(moving_address_src)
async def _run_proactive_reconnect_without_locking(
self,
moving_address_src: str | None = None,
) -> None:
"""
Mark and disconnect matching connections while the caller holds the pool lock.
This helper intentionally does not acquire the pool lock so it can be
reused by larger atomic operations that already hold the non-reentrant
`asyncio.Lock`.
"""
for conn in self._get_in_use_connections():
if self._should_update_connection(
conn, "connected_address", moving_address_src
):
conn.mark_for_reconnect()
free_connections = [
conn
for conn in self._get_free_connections()
if self._should_update_connection(
conn, "connected_address", moving_address_src
)
]
await self._disconnect_connections(free_connections)
async def cleanup_moving_notification(
self,
notification_hash: int,
reset_relaxed_timeout: bool,
reset_host_address: bool,
) -> None:
"""
Revert MOVING pool state atomically after the notification TTL.
Future connection kwargs and existing connection state must be cleaned
up in the same critical section. Splitting the cleanup lets an
acquire/release interleave, which can leave stale MOVING state or undo a
newer overlapping MOVING notification.
"""
async with self._get_pool_lock():
kwargs = _build_moving_cleanup_connection_kwargs(
self.connection_kwargs, notification_hash
)
if kwargs is not None:
self.update_connection_kwargs(**kwargs)
self._update_connections_settings_without_locking(
relaxed_timeout=-1,
state=MaintenanceState.NONE,
maintenance_notification_hash=None,
matching_notification_hash=notification_hash,
matching_pattern="notification_hash",
update_notification_hash=True,
reset_relaxed_timeout=reset_relaxed_timeout,
reset_host_address=reset_host_address,
include_free_connections=True,
)
async def _disconnect_connections(
self, connections: Iterable["AbstractConnection"]
) -> None:
connections = tuple(connections)
if not connections:
return
results = await asyncio.gather(
*(connection.disconnect() for connection in connections),
return_exceptions=True,
)
exc = next(
(result for result in results if isinstance(result, BaseException)), None
)
if exc:
raise exc
def _set_in_maintenance(self, in_maintenance: bool) -> None:
"""Flip the pool's maintenance flag if it exposes one (BlockingConnectionPool)."""
set_in_maintenance = getattr(self, "set_in_maintenance", None)
if callable(set_in_maintenance):
set_in_maintenance(in_maintenance)
[docs]class ConnectionPool(
AsyncMaintNotificationsAbstractConnectionPool, ConnectionPoolInterface
):
"""
Create a connection pool. ``If max_connections`` is set, then this
object raises :py:class:`~redis.ConnectionError` when the pool's
limit is reached.
By default, TCP connections are created unless ``connection_class``
is specified. Use :py:class:`~redis.UnixDomainSocketConnection` for
unix sockets.
:py:class:`~redis.SSLConnection` can be used for SSL enabled connections.
Any additional keyword arguments are passed to the constructor of
``connection_class``.
"""
[docs] @classmethod
def from_url(cls: Type[_CP], url: str, **kwargs) -> _CP:
"""
Return a connection pool configured from the given URL.
For example::
redis://[[username]:[password]]@localhost:6379/0
rediss://[[username]:[password]]@localhost:6379/0
unix://[username@]/path/to/socket.sock?db=0[&password=password]
Three URL schemes are supported:
- `redis://` creates a TCP socket connection. See more at:
<https://www.iana.org/assignments/uri-schemes/prov/redis>
- `rediss://` creates a SSL wrapped TCP socket connection. See more at:
<https://www.iana.org/assignments/uri-schemes/prov/rediss>
- ``unix://``: creates a Unix Domain Socket connection.
The username, password, hostname and path are passed through
urllib.parse.unquote in order to replace any percent-encoded values
with their corresponding characters. Querystring values are decoded
by urllib.parse.parse_qs and are not unquoted again.
There are several ways to specify a database number. The first value
found will be used:
1. A ``db`` querystring option, e.g. redis://localhost?db=0
2. If using the redis:// or rediss:// schemes, the path argument
of the url, e.g. redis://localhost/0
3. A ``db`` keyword argument to this function.
If none of these options are specified, the default db=0 is used.
All querystring options are cast to their appropriate Python types.
Boolean arguments can be specified with string values "True"/"False"
or "Yes"/"No". Values that cannot be properly cast cause a
``ValueError`` to be raised. Once parsed, the querystring arguments
and keyword arguments are passed to the ``ConnectionPool``'s
class initializer. In the case of conflicting arguments, querystring
arguments always win.
"""
url_options = parse_url(url)
kwargs.update(url_options)
return cls(**kwargs)
def __init__(
self,
connection_class: Type[AbstractConnection] = Connection,
max_connections: Optional[int] = None,
maint_notifications_config: MaintNotificationsConfig | None = None,
**connection_kwargs,
):
max_connections = max_connections or 100
if not isinstance(max_connections, int) or max_connections < 0:
raise ValueError('"max_connections" must be a positive integer')
self.connection_class = connection_class
self._connection_kwargs = connection_kwargs
self.max_connections = max_connections
# Resolve the HIMPORT registry. A pre-built ``himport_registry`` (shared, e.g.
# from the cluster client) takes precedence; otherwise build a fresh empty one.
# A registry always exists so runtime ``himport_prepare`` mutates a single object
# every connection already shares. The object stays in ``connection_kwargs`` so
# it reaches every connection. It is injected unconditionally (like other
# auto-added pool kwargs), so a custom ``connection_class`` must accept
# ``**kwargs`` (or a ``himport_registry`` parameter), as built-ins do.
himport_registry = connection_kwargs.get("himport_registry")
if himport_registry is None:
himport_registry = HImportRegistry()
connection_kwargs["himport_registry"] = himport_registry
self.himport_registry = himport_registry
self._available_connections: List[AbstractConnection] = []
self._in_use_connections: Set[AbstractConnection] = set()
self.encoder_class = self.connection_kwargs.get("encoder_class", Encoder)
self._lock = asyncio.Lock()
self._event_dispatcher = self.connection_kwargs.get("event_dispatcher", None)
if self._event_dispatcher is None:
self._event_dispatcher = EventDispatcher()
AsyncMaintNotificationsAbstractConnectionPool.__init__(
self,
maint_notifications_config=maint_notifications_config,
**connection_kwargs,
)
# Keys that should be redacted in __repr__ to avoid exposing sensitive information
SENSITIVE_REPR_KEYS = frozenset(
{
"password",
"username",
"ssl_password",
"credential_provider",
}
)
# Internal plumbing kwargs omitted from __repr__ (not user-facing config).
OMIT_REPR_KEYS = frozenset({"himport_registry"})
def __repr__(self):
conn_kwargs = ",".join(
[
f"{k}={'<REDACTED>' if k in self.SENSITIVE_REPR_KEYS else v}"
for k, v in self.connection_kwargs.items()
if k not in self.OMIT_REPR_KEYS
]
)
return (
f"<{self.__class__.__module__}.{self.__class__.__name__}"
f"(<{self.connection_class.__module__}.{self.connection_class.__name__}"
f"({conn_kwargs})>)>"
)
@property
def connection_kwargs(self) -> dict[str, Any]:
return self._connection_kwargs
@connection_kwargs.setter
def connection_kwargs(self, value: dict[str, Any]) -> None:
self._connection_kwargs = value
def _get_pool_lock(self) -> asyncio.Lock:
return self._lock
def _get_free_connections(self) -> Iterable[AbstractConnection]:
return self._available_connections
def _get_in_use_connections(self) -> Iterable[AbstractConnection]:
return self._in_use_connections
[docs] def get_protocol(self):
"""
Returns:
The RESP protocol version, or ``None`` if the protocol is not specified,
in which case the server default will be used.
"""
return self.connection_kwargs.get("protocol", None)
def reset(self):
# Record metrics for connections being removed before clearing
# (only if attributes exist - they won't during __init__)
if hasattr(self, "_available_connections") and hasattr(
self, "_in_use_connections"
):
idle_count = len(self._available_connections)
in_use_count = len(self._in_use_connections)
if idle_count > 0 or in_use_count > 0:
pool_name = get_pool_name(self)
# Note: Using sync version since reset() is sync
from redis.observability.recorder import (
record_connection_count as sync_record_connection_count,
)
if idle_count > 0:
sync_record_connection_count(
pool_name=pool_name,
connection_state=ConnectionState.IDLE,
counter=-idle_count,
)
if in_use_count > 0:
sync_record_connection_count(
pool_name=pool_name,
connection_state=ConnectionState.USED,
counter=-in_use_count,
)
self._available_connections = []
self._in_use_connections = weakref.WeakSet()
def __del__(self) -> None:
"""Clean up connection pool and record metrics when garbage collected."""
try:
if not hasattr(self, "_available_connections") or not hasattr(
self, "_in_use_connections"
):
return
idle_count = len(self._available_connections)
in_use_count = len(self._in_use_connections)
if idle_count > 0 or in_use_count > 0:
pool_name = get_pool_name(self)
# Note: Using sync version since __del__ is sync
from redis.observability.recorder import (
record_connection_count as sync_record_connection_count,
)
if idle_count > 0:
sync_record_connection_count(
pool_name=pool_name,
connection_state=ConnectionState.IDLE,
counter=-idle_count,
)
if in_use_count > 0:
sync_record_connection_count(
pool_name=pool_name,
connection_state=ConnectionState.USED,
counter=-in_use_count,
)
except Exception:
pass
[docs] def can_get_connection(self) -> bool:
"""Return True if a connection can be retrieved from the pool."""
return (
self._available_connections
or len(self._in_use_connections) < self.max_connections
)
[docs] @deprecated_args(
args_to_warn=["*"],
reason="Use get_connection() without args instead",
version="5.3.0",
)
async def get_connection(self, command_name=None, *keys, **options):
"""Get a connected connection from the pool"""
# Track connection count before to detect if a new connection is created
async with self._lock:
connections_before = len(self._available_connections) + len(
self._in_use_connections
)
start_time_created = time.monotonic()
connection = self.get_available_connection()
connections_after = len(self._available_connections) + len(
self._in_use_connections
)
is_created = connections_after > connections_before
# Record state transition for observability
# This ensures counters stay balanced if ensure_connection() fails and release() is called
pool_name = get_pool_name(self)
if is_created:
# New connection created and acquired: just USED +1
await record_connection_count(
pool_name=pool_name,
connection_state=ConnectionState.USED,
counter=1,
)
else:
# Existing connection acquired from pool: IDLE -> USED
await record_connection_count(
pool_name=pool_name,
connection_state=ConnectionState.IDLE,
counter=-1,
)
await record_connection_count(
pool_name=pool_name,
connection_state=ConnectionState.USED,
counter=1,
)
# We now perform the connection check outside of the lock.
try:
await self.ensure_connection(connection)
if is_created:
await record_connection_create_time(
connection_pool=self,
duration_seconds=time.monotonic() - start_time_created,
)
return connection
except BaseException:
await self.release(connection)
raise
[docs] def get_available_connection(self):
"""Get a connection from the pool, without making sure it is connected"""
try:
connection = self._available_connections.pop()
except IndexError:
if len(self._in_use_connections) >= self.max_connections:
raise MaxConnectionsError("Too many connections") from None
connection = self.make_connection()
self._in_use_connections.add(connection)
return connection
[docs] def get_encoder(self):
"""Return an encoder based on encoding settings"""
kwargs = self.connection_kwargs
return self.encoder_class(
encoding=kwargs.get("encoding", "utf-8"),
encoding_errors=kwargs.get("encoding_errors", "strict"),
decode_responses=kwargs.get("decode_responses", False),
)
[docs] def make_connection(self):
"""Create a new connection. Can be overridden by child classes."""
# Note: We don't record IDLE here because async uses a sync make_connection
# but async record_connection_count. The recording is handled in get_connection.
return self.connection_class(**self.connection_kwargs)
[docs] async def ensure_connection(self, connection: AbstractConnection):
"""Ensure that the connection object is connected and valid"""
await connection.connect()
# connections that the pool provides should be ready to send
# a command. if not, the connection was either returned to the
# pool before all data has been read or the socket has been
# closed. either way, reconnect and verify everything is good.
try:
if await connection.can_read() and not self.maint_notifications_enabled():
raise ConnectionError("Connection has data") from None
except (ConnectionError, TimeoutError, OSError):
await connection.disconnect()
await connection.connect()
if await connection.can_read() and not self.maint_notifications_enabled():
raise ConnectionError("Connection not ready") from None
[docs] async def release(self, connection: AbstractConnection):
"""Releases the connection back to the pool"""
# Connections should always be returned to the correct pool,
# not doing so is an error that will cause an exception here.
async with self._lock:
self._in_use_connections.remove(connection)
if connection.should_reconnect():
await connection.disconnect()
self._available_connections.append(connection)
await self._event_dispatcher.dispatch_async(
AsyncAfterConnectionReleasedEvent(connection)
)
# Record state transition: USED -> IDLE
pool_name = get_pool_name(self)
await record_connection_count(
pool_name=pool_name,
connection_state=ConnectionState.USED,
counter=-1,
)
await record_connection_count(
pool_name=pool_name,
connection_state=ConnectionState.IDLE,
counter=1,
)
[docs] async def disconnect(self, inuse_connections: bool = True):
"""
Disconnects connections in the pool
If ``inuse_connections`` is True, disconnect connections that are
current in use, potentially by other tasks. Otherwise only disconnect
connections that are idle in the pool.
"""
if inuse_connections:
connections: Iterable[AbstractConnection] = chain(
self._available_connections, self._in_use_connections
)
else:
connections = self._available_connections
resp = await asyncio.gather(
*(connection.disconnect() for connection in connections),
return_exceptions=True,
)
exc = next((r for r in resp if isinstance(r, BaseException)), None)
if exc:
raise exc
[docs] async def update_active_connections_for_reconnect(self):
"""
Mark all active connections for reconnect.
"""
async with self._lock:
for conn in self._in_use_connections:
conn.mark_for_reconnect()
[docs] async def aclose(self) -> None:
"""Close the pool, disconnecting all connections"""
await self._on_close()
await self.disconnect()
async def __aenter__(self: _CP) -> _CP:
return self
async def __aexit__(self, exc_type, exc_value, traceback) -> None:
await self.aclose()
def set_retry(self, retry: "Retry") -> None:
for conn in self._available_connections:
conn.retry = retry
for conn in self._in_use_connections:
conn.retry = retry
async def re_auth_callback(self, token: TokenInterface):
async with self._lock:
for conn in self._available_connections:
await conn.retry.call_with_retry(
lambda: conn.send_command(
"AUTH", token.try_get("oid"), token.get_value()
),
lambda error: self._mock(error),
)
await conn.retry.call_with_retry(
lambda: conn.read_response(), lambda error: self._mock(error)
)
for conn in self._in_use_connections:
conn.set_re_auth_token(token)
async def _mock(self, error: RedisError):
"""
Dummy functions, needs to be passed as error callback to retry object.
:param error:
:return:
"""
pass
[docs] def get_connection_count(self) -> List[tuple[int, dict]]:
"""
Returns a connection count (both idle and in use).
"""
attributes = AttributeBuilder.build_base_attributes()
attributes[DB_CLIENT_CONNECTION_POOL_NAME] = get_pool_name(self)
free_connections_attributes = attributes.copy()
in_use_connections_attributes = attributes.copy()
free_connections_attributes[DB_CLIENT_CONNECTION_STATE] = (
ConnectionState.IDLE.value
)
in_use_connections_attributes[DB_CLIENT_CONNECTION_STATE] = (
ConnectionState.USED.value
)
return [
(len(self._available_connections), free_connections_attributes),
(len(self._in_use_connections), in_use_connections_attributes),
]
class BlockingConnectionPool(ConnectionPool):
"""
A blocking connection pool::
>>> from redis.asyncio import Redis, BlockingConnectionPool
>>> client = Redis.from_pool(BlockingConnectionPool())
It performs the same function as the default
:py:class:`~redis.asyncio.ConnectionPool` implementation, in that,
it maintains a pool of reusable connections that can be shared by
multiple async redis clients.
The difference is that, in the event that a client tries to get a
connection from the pool when all of connections are in use, rather than
raising a :py:class:`~redis.ConnectionError` (as the default
:py:class:`~redis.asyncio.ConnectionPool` implementation does), it
blocks the current `Task` for a specified number of seconds until
a connection becomes available.
Use ``max_connections`` to increase / decrease the pool size::
>>> pool = BlockingConnectionPool(max_connections=10)
Use ``timeout`` to tell it either how many seconds to wait for a connection
to become available, or to block forever:
>>> # Block forever.
>>> pool = BlockingConnectionPool(timeout=None)
>>> # Raise a ``ConnectionError`` after five seconds if a connection is
>>> # not available.
>>> pool = BlockingConnectionPool(timeout=5)
"""
def __init__(
self,
max_connections: int = 50,
timeout: Optional[float] = 20,
connection_class: Type[AbstractConnection] = Connection,
queue_class: Type[asyncio.Queue] = asyncio.LifoQueue, # deprecated
**connection_kwargs,
):
super().__init__(
connection_class=connection_class,
max_connections=max_connections,
**connection_kwargs,
)
self._condition = asyncio.Condition()
self.timeout = timeout
self._in_maintenance = False
def set_in_maintenance(self, in_maintenance: bool) -> None:
"""
Toggle the pool's maintenance mode.
While maintenance mode is on, ``get_connection`` and ``release``
serialize their pool mutations through ``self._lock`` so they cannot
interleave with a MOVING notification handler that is currently
rewriting pool state under the same lock. Outside of maintenance the
mutations skip the lock, since their critical sections are pure-Python
and already atomic under asyncio's single-threaded scheduling.
"""
self._in_maintenance = in_maintenance
@contextlib.asynccontextmanager
async def _maybe_pool_lock(self) -> AsyncIterator[None]:
if self._in_maintenance:
async with self._lock:
yield
else:
yield
@deprecated_args(
args_to_warn=["*"],
reason="Use get_connection() without args instead",
version="5.3.0",
)
async def get_connection(self, command_name=None, *keys, **options):
"""Gets a connection from the pool, blocking until one is available"""
# Start timing for wait time observability
start_time_acquired = time.monotonic()
try:
async with self._condition:
async with async_timeout(self.timeout):
await self._condition.wait_for(self.can_get_connection)
async with self._maybe_pool_lock():
# Track connection count before to detect if a new connection is created
connections_before = len(self._available_connections) + len(
self._in_use_connections
)
start_time_created = time.monotonic()
connection = super().get_available_connection()
connections_after = len(self._available_connections) + len(
self._in_use_connections
)
is_created = connections_after > connections_before
except asyncio.TimeoutError as err:
raise ConnectionError("No connection available.") from err
# We now perform the connection check outside of the lock.
try:
await self.ensure_connection(connection)
if is_created:
await record_connection_create_time(
connection_pool=self,
duration_seconds=time.monotonic() - start_time_created,
)
await record_connection_wait_time(
pool_name=get_pool_name(self),
duration_seconds=time.monotonic() - start_time_acquired,
)
return connection
except BaseException:
await self.release(connection)
raise
async def release(self, connection: AbstractConnection):
"""Releases the connection back to the pool."""
async with self._condition:
await super().release(connection)
self._condition.notify()