Coverage for /pythoncovmergedfiles/medio/medio/usr/local/lib/python3.11/site-packages/redis/cluster.py: 21%
Shortcuts on this page
r m x toggle line displays
j k next/prev highlighted chunk
0 (zero) top of page
1 (one) first highlighted chunk
Shortcuts on this page
r m x toggle line displays
j k next/prev highlighted chunk
0 (zero) top of page
1 (one) first highlighted chunk
1import random
2import socket
3import sys
4import threading
5import time
6from abc import ABC, abstractmethod
7from collections import OrderedDict
8from copy import copy
9from enum import Enum
10from itertools import chain
11from typing import Any, Callable, Dict, List, Optional, Set, Tuple, Union
13from redis._parsers import CommandsParser, Encoder
14from redis._parsers.commands import CommandPolicies, RequestPolicy, ResponsePolicy
15from redis._parsers.helpers import parse_scan
16from redis.backoff import ExponentialWithJitterBackoff, NoBackoff
17from redis.cache import CacheConfig, CacheFactory, CacheFactoryInterface, CacheInterface
18from redis.client import EMPTY_RESPONSE, CaseInsensitiveDict, PubSub, Redis
19from redis.commands import READ_COMMANDS, RedisClusterCommands
20from redis.commands.helpers import list_or_args
21from redis.commands.policies import PolicyResolver, StaticPolicyResolver
22from redis.connection import (
23 Connection,
24 ConnectionPool,
25 parse_url,
26)
27from redis.crc import REDIS_CLUSTER_HASH_SLOTS, key_slot
28from redis.event import (
29 AfterPooledConnectionsInstantiationEvent,
30 AfterPubSubConnectionInstantiationEvent,
31 ClientType,
32 EventDispatcher,
33)
34from redis.exceptions import (
35 AskError,
36 AuthenticationError,
37 ClusterDownError,
38 ClusterError,
39 ConnectionError,
40 CrossSlotTransactionError,
41 DataError,
42 ExecAbortError,
43 InvalidPipelineStack,
44 MaxConnectionsError,
45 MovedError,
46 RedisClusterException,
47 RedisError,
48 ResponseError,
49 SlotNotCoveredError,
50 TimeoutError,
51 TryAgainError,
52 WatchError,
53)
54from redis.lock import Lock
55from redis.maint_notifications import MaintNotificationsConfig
56from redis.retry import Retry
57from redis.utils import (
58 deprecated_args,
59 dict_merge,
60 list_keys_to_dict,
61 merge_result,
62 safe_str,
63 str_if_bytes,
64 truncate_text,
65)
68def get_node_name(host: str, port: Union[str, int]) -> str:
69 return f"{host}:{port}"
72@deprecated_args(
73 allowed_args=["redis_node"],
74 reason="Use get_connection(redis_node) instead",
75 version="5.3.0",
76)
77def get_connection(redis_node: Redis, *args, **options) -> Connection:
78 return redis_node.connection or redis_node.connection_pool.get_connection()
81def parse_scan_result(command, res, **options):
82 cursors = {}
83 ret = []
84 for node_name, response in res.items():
85 cursor, r = parse_scan(response, **options)
86 cursors[node_name] = cursor
87 ret += r
89 return cursors, ret
92def parse_pubsub_numsub(command, res, **options):
93 numsub_d = OrderedDict()
94 for numsub_tups in res.values():
95 for channel, numsubbed in numsub_tups:
96 try:
97 numsub_d[channel] += numsubbed
98 except KeyError:
99 numsub_d[channel] = numsubbed
101 ret_numsub = [(channel, numsub) for channel, numsub in numsub_d.items()]
102 return ret_numsub
105def parse_cluster_slots(
106 resp: Any, **options: Any
107) -> Dict[Tuple[int, int], Dict[str, Any]]:
108 current_host = options.get("current_host", "")
110 def fix_server(*args: Any) -> Tuple[str, Any]:
111 return str_if_bytes(args[0]) or current_host, args[1]
113 slots = {}
114 for slot in resp:
115 start, end, primary = slot[:3]
116 replicas = slot[3:]
117 slots[start, end] = {
118 "primary": fix_server(*primary),
119 "replicas": [fix_server(*replica) for replica in replicas],
120 }
122 return slots
125def parse_cluster_shards(resp, **options):
126 """
127 Parse CLUSTER SHARDS response.
128 """
129 if isinstance(resp[0], dict):
130 return resp
131 shards = []
132 for x in resp:
133 shard = {"slots": [], "nodes": []}
134 for i in range(0, len(x[1]), 2):
135 shard["slots"].append((x[1][i], (x[1][i + 1])))
136 nodes = x[3]
137 for node in nodes:
138 dict_node = {}
139 for i in range(0, len(node), 2):
140 dict_node[node[i]] = node[i + 1]
141 shard["nodes"].append(dict_node)
142 shards.append(shard)
144 return shards
147def parse_cluster_myshardid(resp, **options):
148 """
149 Parse CLUSTER MYSHARDID response.
150 """
151 return resp.decode("utf-8")
154PRIMARY = "primary"
155REPLICA = "replica"
156SLOT_ID = "slot-id"
158REDIS_ALLOWED_KEYS = (
159 "connection_class",
160 "connection_pool",
161 "connection_pool_class",
162 "client_name",
163 "credential_provider",
164 "db",
165 "decode_responses",
166 "encoding",
167 "encoding_errors",
168 "host",
169 "lib_name",
170 "lib_version",
171 "max_connections",
172 "nodes_flag",
173 "redis_connect_func",
174 "password",
175 "port",
176 "timeout",
177 "queue_class",
178 "retry",
179 "retry_on_timeout",
180 "protocol",
181 "socket_connect_timeout",
182 "socket_keepalive",
183 "socket_keepalive_options",
184 "socket_timeout",
185 "ssl",
186 "ssl_ca_certs",
187 "ssl_ca_data",
188 "ssl_ca_path",
189 "ssl_certfile",
190 "ssl_cert_reqs",
191 "ssl_include_verify_flags",
192 "ssl_exclude_verify_flags",
193 "ssl_keyfile",
194 "ssl_password",
195 "ssl_check_hostname",
196 "unix_socket_path",
197 "username",
198 "cache",
199 "cache_config",
200)
201KWARGS_DISABLED_KEYS = ("host", "port", "retry")
204def cleanup_kwargs(**kwargs):
205 """
206 Remove unsupported or disabled keys from kwargs
207 """
208 connection_kwargs = {
209 k: v
210 for k, v in kwargs.items()
211 if k in REDIS_ALLOWED_KEYS and k not in KWARGS_DISABLED_KEYS
212 }
214 return connection_kwargs
217class AbstractRedisCluster:
218 RedisClusterRequestTTL = 16
220 PRIMARIES = "primaries"
221 REPLICAS = "replicas"
222 ALL_NODES = "all"
223 RANDOM = "random"
224 DEFAULT_NODE = "default-node"
226 NODE_FLAGS = {PRIMARIES, REPLICAS, ALL_NODES, RANDOM, DEFAULT_NODE}
228 COMMAND_FLAGS = dict_merge(
229 list_keys_to_dict(
230 [
231 "ACL CAT",
232 "ACL DELUSER",
233 "ACL DRYRUN",
234 "ACL GENPASS",
235 "ACL GETUSER",
236 "ACL HELP",
237 "ACL LIST",
238 "ACL LOG",
239 "ACL LOAD",
240 "ACL SAVE",
241 "ACL SETUSER",
242 "ACL USERS",
243 "ACL WHOAMI",
244 "AUTH",
245 "CLIENT LIST",
246 "CLIENT SETINFO",
247 "CLIENT SETNAME",
248 "CLIENT GETNAME",
249 "CONFIG SET",
250 "CONFIG REWRITE",
251 "CONFIG RESETSTAT",
252 "TIME",
253 "PUBSUB CHANNELS",
254 "PUBSUB NUMPAT",
255 "PUBSUB NUMSUB",
256 "PUBSUB SHARDCHANNELS",
257 "PUBSUB SHARDNUMSUB",
258 "PING",
259 "INFO",
260 "SHUTDOWN",
261 "KEYS",
262 "DBSIZE",
263 "BGSAVE",
264 "SLOWLOG GET",
265 "SLOWLOG LEN",
266 "SLOWLOG RESET",
267 "WAIT",
268 "WAITAOF",
269 "SAVE",
270 "MEMORY PURGE",
271 "MEMORY MALLOC-STATS",
272 "MEMORY STATS",
273 "LASTSAVE",
274 "CLIENT TRACKINGINFO",
275 "CLIENT PAUSE",
276 "CLIENT UNPAUSE",
277 "CLIENT UNBLOCK",
278 "CLIENT ID",
279 "CLIENT REPLY",
280 "CLIENT GETREDIR",
281 "CLIENT INFO",
282 "CLIENT KILL",
283 "READONLY",
284 "CLUSTER INFO",
285 "CLUSTER MEET",
286 "CLUSTER MYSHARDID",
287 "CLUSTER NODES",
288 "CLUSTER REPLICAS",
289 "CLUSTER RESET",
290 "CLUSTER SET-CONFIG-EPOCH",
291 "CLUSTER SLOTS",
292 "CLUSTER SHARDS",
293 "CLUSTER COUNT-FAILURE-REPORTS",
294 "CLUSTER KEYSLOT",
295 "COMMAND",
296 "COMMAND COUNT",
297 "COMMAND LIST",
298 "COMMAND GETKEYS",
299 "CONFIG GET",
300 "DEBUG",
301 "RANDOMKEY",
302 "READONLY",
303 "READWRITE",
304 "TIME",
305 "TFUNCTION LOAD",
306 "TFUNCTION DELETE",
307 "TFUNCTION LIST",
308 "TFCALL",
309 "TFCALLASYNC",
310 "LATENCY HISTORY",
311 "LATENCY LATEST",
312 "LATENCY RESET",
313 "MODULE LIST",
314 "MODULE LOAD",
315 "MODULE UNLOAD",
316 "MODULE LOADEX",
317 ],
318 DEFAULT_NODE,
319 ),
320 list_keys_to_dict(
321 [
322 "FLUSHALL",
323 "FLUSHDB",
324 "FUNCTION DELETE",
325 "FUNCTION FLUSH",
326 "FUNCTION LIST",
327 "FUNCTION LOAD",
328 "FUNCTION RESTORE",
329 "SCAN",
330 "SCRIPT EXISTS",
331 "SCRIPT FLUSH",
332 "SCRIPT LOAD",
333 ],
334 PRIMARIES,
335 ),
336 list_keys_to_dict(["FUNCTION DUMP"], RANDOM),
337 list_keys_to_dict(
338 [
339 "CLUSTER COUNTKEYSINSLOT",
340 "CLUSTER DELSLOTS",
341 "CLUSTER DELSLOTSRANGE",
342 "CLUSTER GETKEYSINSLOT",
343 "CLUSTER SETSLOT",
344 ],
345 SLOT_ID,
346 ),
347 )
349 SEARCH_COMMANDS = (
350 [
351 "FT.CREATE",
352 "FT.SEARCH",
353 "FT.AGGREGATE",
354 "FT.EXPLAIN",
355 "FT.EXPLAINCLI",
356 "FT,PROFILE",
357 "FT.ALTER",
358 "FT.DROPINDEX",
359 "FT.ALIASADD",
360 "FT.ALIASUPDATE",
361 "FT.ALIASDEL",
362 "FT.TAGVALS",
363 "FT.SUGADD",
364 "FT.SUGGET",
365 "FT.SUGDEL",
366 "FT.SUGLEN",
367 "FT.SYNUPDATE",
368 "FT.SYNDUMP",
369 "FT.SPELLCHECK",
370 "FT.DICTADD",
371 "FT.DICTDEL",
372 "FT.DICTDUMP",
373 "FT.INFO",
374 "FT._LIST",
375 "FT.CONFIG",
376 "FT.ADD",
377 "FT.DEL",
378 "FT.DROP",
379 "FT.GET",
380 "FT.MGET",
381 "FT.SYNADD",
382 ],
383 )
385 CLUSTER_COMMANDS_RESPONSE_CALLBACKS = {
386 "CLUSTER SLOTS": parse_cluster_slots,
387 "CLUSTER SHARDS": parse_cluster_shards,
388 "CLUSTER MYSHARDID": parse_cluster_myshardid,
389 }
391 RESULT_CALLBACKS = dict_merge(
392 list_keys_to_dict(["PUBSUB NUMSUB", "PUBSUB SHARDNUMSUB"], parse_pubsub_numsub),
393 list_keys_to_dict(
394 ["PUBSUB NUMPAT"], lambda command, res: sum(list(res.values()))
395 ),
396 list_keys_to_dict(
397 ["KEYS", "PUBSUB CHANNELS", "PUBSUB SHARDCHANNELS"], merge_result
398 ),
399 list_keys_to_dict(
400 [
401 "PING",
402 "CONFIG SET",
403 "CONFIG REWRITE",
404 "CONFIG RESETSTAT",
405 "CLIENT SETNAME",
406 "BGSAVE",
407 "SLOWLOG RESET",
408 "SAVE",
409 "MEMORY PURGE",
410 "CLIENT PAUSE",
411 "CLIENT UNPAUSE",
412 ],
413 lambda command, res: all(res.values()) if isinstance(res, dict) else res,
414 ),
415 list_keys_to_dict(
416 ["DBSIZE", "WAIT"],
417 lambda command, res: sum(res.values()) if isinstance(res, dict) else res,
418 ),
419 list_keys_to_dict(
420 ["CLIENT UNBLOCK"], lambda command, res: 1 if sum(res.values()) > 0 else 0
421 ),
422 list_keys_to_dict(["SCAN"], parse_scan_result),
423 list_keys_to_dict(
424 ["SCRIPT LOAD"], lambda command, res: list(res.values()).pop()
425 ),
426 list_keys_to_dict(
427 ["SCRIPT EXISTS"], lambda command, res: [all(k) for k in zip(*res.values())]
428 ),
429 list_keys_to_dict(["SCRIPT FLUSH"], lambda command, res: all(res.values())),
430 )
432 ERRORS_ALLOW_RETRY = (
433 ConnectionError,
434 TimeoutError,
435 ClusterDownError,
436 SlotNotCoveredError,
437 )
439 def replace_default_node(self, target_node: "ClusterNode" = None) -> None:
440 """Replace the default cluster node.
441 A random cluster node will be chosen if target_node isn't passed, and primaries
442 will be prioritized. The default node will not be changed if there are no other
443 nodes in the cluster.
445 Args:
446 target_node (ClusterNode, optional): Target node to replace the default
447 node. Defaults to None.
448 """
449 if target_node:
450 self.nodes_manager.default_node = target_node
451 else:
452 curr_node = self.get_default_node()
453 primaries = [node for node in self.get_primaries() if node != curr_node]
454 if primaries:
455 # Choose a primary if the cluster contains different primaries
456 self.nodes_manager.default_node = random.choice(primaries)
457 else:
458 # Otherwise, choose a primary if the cluster contains different primaries
459 replicas = [node for node in self.get_replicas() if node != curr_node]
460 if replicas:
461 self.nodes_manager.default_node = random.choice(replicas)
464class RedisCluster(AbstractRedisCluster, RedisClusterCommands):
465 @classmethod
466 def from_url(cls, url: str, **kwargs: Any) -> "RedisCluster":
467 """
468 Return a Redis client object configured from the given URL
470 For example::
472 redis://[[username]:[password]]@localhost:6379/0
473 rediss://[[username]:[password]]@localhost:6379/0
474 unix://[username@]/path/to/socket.sock?db=0[&password=password]
476 Three URL schemes are supported:
478 - `redis://` creates a TCP socket connection. See more at:
479 <https://www.iana.org/assignments/uri-schemes/prov/redis>
480 - `rediss://` creates a SSL wrapped TCP socket connection. See more at:
481 <https://www.iana.org/assignments/uri-schemes/prov/rediss>
482 - ``unix://``: creates a Unix Domain Socket connection.
484 The username, password, hostname, path and all querystring values
485 are passed through urllib.parse.unquote in order to replace any
486 percent-encoded values with their corresponding characters.
488 There are several ways to specify a database number. The first value
489 found will be used:
491 1. A ``db`` querystring option, e.g. redis://localhost?db=0
492 2. If using the redis:// or rediss:// schemes, the path argument
493 of the url, e.g. redis://localhost/0
494 3. A ``db`` keyword argument to this function.
496 If none of these options are specified, the default db=0 is used.
498 All querystring options are cast to their appropriate Python types.
499 Boolean arguments can be specified with string values "True"/"False"
500 or "Yes"/"No". Values that cannot be properly cast cause a
501 ``ValueError`` to be raised. Once parsed, the querystring arguments
502 and keyword arguments are passed to the ``ConnectionPool``'s
503 class initializer. In the case of conflicting arguments, querystring
504 arguments always win.
506 """
507 return cls(url=url, **kwargs)
509 @deprecated_args(
510 args_to_warn=["read_from_replicas"],
511 reason="Please configure the 'load_balancing_strategy' instead",
512 version="5.3.0",
513 )
514 @deprecated_args(
515 args_to_warn=[
516 "cluster_error_retry_attempts",
517 ],
518 reason="Please configure the 'retry' object instead",
519 version="6.0.0",
520 )
521 def __init__(
522 self,
523 host: Optional[str] = None,
524 port: int = 6379,
525 startup_nodes: Optional[List["ClusterNode"]] = None,
526 cluster_error_retry_attempts: int = 3,
527 retry: Optional["Retry"] = None,
528 require_full_coverage: bool = True,
529 reinitialize_steps: int = 5,
530 read_from_replicas: bool = False,
531 load_balancing_strategy: Optional["LoadBalancingStrategy"] = None,
532 dynamic_startup_nodes: bool = True,
533 url: Optional[str] = None,
534 address_remap: Optional[Callable[[Tuple[str, int]], Tuple[str, int]]] = None,
535 cache: Optional[CacheInterface] = None,
536 cache_config: Optional[CacheConfig] = None,
537 event_dispatcher: Optional[EventDispatcher] = None,
538 policy_resolver: PolicyResolver = StaticPolicyResolver(),
539 **kwargs,
540 ):
541 """
542 Initialize a new RedisCluster client.
544 :param startup_nodes:
545 List of nodes from which initial bootstrapping can be done
546 :param host:
547 Can be used to point to a startup node
548 :param port:
549 Can be used to point to a startup node
550 :param require_full_coverage:
551 When set to False (default value): the client will not require a
552 full coverage of the slots. However, if not all slots are covered,
553 and at least one node has 'cluster-require-full-coverage' set to
554 'yes,' the server will throw a ClusterDownError for some key-based
555 commands. See -
556 https://redis.io/topics/cluster-tutorial#redis-cluster-configuration-parameters
557 When set to True: all slots must be covered to construct the
558 cluster client. If not all slots are covered, RedisClusterException
559 will be thrown.
560 :param read_from_replicas:
561 @deprecated - please use load_balancing_strategy instead
562 Enable read from replicas in READONLY mode. You can read possibly
563 stale data.
564 When set to true, read commands will be assigned between the
565 primary and its replications in a Round-Robin manner.
566 :param load_balancing_strategy:
567 Enable read from replicas in READONLY mode and defines the load balancing
568 strategy that will be used for cluster node selection.
569 The data read from replicas is eventually consistent with the data in primary nodes.
570 :param dynamic_startup_nodes:
571 Set the RedisCluster's startup nodes to all of the discovered nodes.
572 If true (default value), the cluster's discovered nodes will be used to
573 determine the cluster nodes-slots mapping in the next topology refresh.
574 It will remove the initial passed startup nodes if their endpoints aren't
575 listed in the CLUSTER SLOTS output.
576 If you use dynamic DNS endpoints for startup nodes but CLUSTER SLOTS lists
577 specific IP addresses, it is best to set it to false.
578 :param cluster_error_retry_attempts:
579 @deprecated - Please configure the 'retry' object instead
580 In case 'retry' object is set - this argument is ignored!
582 Number of times to retry before raising an error when
583 :class:`~.TimeoutError` or :class:`~.ConnectionError`, :class:`~.SlotNotCoveredError` or
584 :class:`~.ClusterDownError` are encountered
585 :param retry:
586 A retry object that defines the retry strategy and the number of
587 retries for the cluster client.
588 In current implementation for the cluster client (starting form redis-py version 6.0.0)
589 the retry object is not yet fully utilized, instead it is used just to determine
590 the number of retries for the cluster client.
591 In the future releases the retry object will be used to handle the cluster client retries!
592 :param reinitialize_steps:
593 Specifies the number of MOVED errors that need to occur before
594 reinitializing the whole cluster topology. If a MOVED error occurs
595 and the cluster does not need to be reinitialized on this current
596 error handling, only the MOVED slot will be patched with the
597 redirected node.
598 To reinitialize the cluster on every MOVED error, set
599 reinitialize_steps to 1.
600 To avoid reinitializing the cluster on moved errors, set
601 reinitialize_steps to 0.
602 :param address_remap:
603 An optional callable which, when provided with an internal network
604 address of a node, e.g. a `(host, port)` tuple, will return the address
605 where the node is reachable. This can be used to map the addresses at
606 which the nodes _think_ they are, to addresses at which a client may
607 reach them, such as when they sit behind a proxy.
609 :**kwargs:
610 Extra arguments that will be sent into Redis instance when created
611 (See Official redis-py doc for supported kwargs - the only limitation
612 is that you can't provide 'retry' object as part of kwargs.
613 [https://github.com/andymccurdy/redis-py/blob/master/redis/client.py])
614 Some kwargs are not supported and will raise a
615 RedisClusterException:
616 - db (Redis do not support database SELECT in cluster mode)
617 """
618 if startup_nodes is None:
619 startup_nodes = []
621 if "db" in kwargs:
622 # Argument 'db' is not possible to use in cluster mode
623 raise RedisClusterException(
624 "Argument 'db' is not possible to use in cluster mode"
625 )
627 if "retry" in kwargs:
628 # Argument 'retry' is not possible to be used in kwargs when in cluster mode
629 # the kwargs are set to the lower level connections to the cluster nodes
630 # and there we provide retry configuration without retries allowed.
631 # The retries should be handled on cluster client level.
632 raise RedisClusterException(
633 "The 'retry' argument cannot be used in kwargs when running in cluster mode."
634 )
636 # Get the startup node/s
637 from_url = False
638 if url is not None:
639 from_url = True
640 url_options = parse_url(url)
641 if "path" in url_options:
642 raise RedisClusterException(
643 "RedisCluster does not currently support Unix Domain "
644 "Socket connections"
645 )
646 if "db" in url_options and url_options["db"] != 0:
647 # Argument 'db' is not possible to use in cluster mode
648 raise RedisClusterException(
649 "A ``db`` querystring option can only be 0 in cluster mode"
650 )
651 kwargs.update(url_options)
652 host = kwargs.get("host")
653 port = kwargs.get("port", port)
654 startup_nodes.append(ClusterNode(host, port))
655 elif host is not None and port is not None:
656 startup_nodes.append(ClusterNode(host, port))
657 elif len(startup_nodes) == 0:
658 # No startup node was provided
659 raise RedisClusterException(
660 "RedisCluster requires at least one node to discover the "
661 "cluster. Please provide one of the followings:\n"
662 "1. host and port, for example:\n"
663 " RedisCluster(host='localhost', port=6379)\n"
664 "2. list of startup nodes, for example:\n"
665 " RedisCluster(startup_nodes=[ClusterNode('localhost', 6379),"
666 " ClusterNode('localhost', 6378)])"
667 )
668 # Update the connection arguments
669 # Whenever a new connection is established, RedisCluster's on_connect
670 # method should be run
671 # If the user passed on_connect function we'll save it and run it
672 # inside the RedisCluster.on_connect() function
673 self.user_on_connect_func = kwargs.pop("redis_connect_func", None)
674 kwargs.update({"redis_connect_func": self.on_connect})
675 kwargs = cleanup_kwargs(**kwargs)
676 if retry:
677 self.retry = retry
678 else:
679 self.retry = Retry(
680 backoff=ExponentialWithJitterBackoff(base=1, cap=10),
681 retries=cluster_error_retry_attempts,
682 )
684 self.encoder = Encoder(
685 kwargs.get("encoding", "utf-8"),
686 kwargs.get("encoding_errors", "strict"),
687 kwargs.get("decode_responses", False),
688 )
689 protocol = kwargs.get("protocol", None)
690 if (cache_config or cache) and protocol not in [3, "3"]:
691 raise RedisError("Client caching is only supported with RESP version 3")
693 self.command_flags = self.__class__.COMMAND_FLAGS.copy()
694 self.node_flags = self.__class__.NODE_FLAGS.copy()
695 self.read_from_replicas = read_from_replicas
696 self.load_balancing_strategy = load_balancing_strategy
697 self.reinitialize_counter = 0
698 self.reinitialize_steps = reinitialize_steps
699 if event_dispatcher is None:
700 self._event_dispatcher = EventDispatcher()
701 else:
702 self._event_dispatcher = event_dispatcher
703 self.startup_nodes = startup_nodes
704 self.nodes_manager = NodesManager(
705 startup_nodes=startup_nodes,
706 from_url=from_url,
707 require_full_coverage=require_full_coverage,
708 dynamic_startup_nodes=dynamic_startup_nodes,
709 address_remap=address_remap,
710 cache=cache,
711 cache_config=cache_config,
712 event_dispatcher=self._event_dispatcher,
713 **kwargs,
714 )
716 self.cluster_response_callbacks = CaseInsensitiveDict(
717 self.__class__.CLUSTER_COMMANDS_RESPONSE_CALLBACKS
718 )
719 self.result_callbacks = CaseInsensitiveDict(self.__class__.RESULT_CALLBACKS)
721 # For backward compatibility, mapping from existing policies to new one
722 self._command_flags_mapping: dict[str, Union[RequestPolicy, ResponsePolicy]] = {
723 self.__class__.RANDOM: RequestPolicy.DEFAULT_KEYLESS,
724 self.__class__.PRIMARIES: RequestPolicy.ALL_SHARDS,
725 self.__class__.ALL_NODES: RequestPolicy.ALL_NODES,
726 self.__class__.REPLICAS: RequestPolicy.ALL_REPLICAS,
727 self.__class__.DEFAULT_NODE: RequestPolicy.DEFAULT_NODE,
728 SLOT_ID: RequestPolicy.DEFAULT_KEYED,
729 }
731 self._policies_callback_mapping: dict[
732 Union[RequestPolicy, ResponsePolicy], Callable
733 ] = {
734 RequestPolicy.DEFAULT_KEYLESS: lambda command_name: [
735 self.get_random_primary_or_all_nodes(command_name)
736 ],
737 RequestPolicy.DEFAULT_KEYED: lambda command,
738 *args: self.get_nodes_from_slot(command, *args),
739 RequestPolicy.DEFAULT_NODE: lambda: [self.get_default_node()],
740 RequestPolicy.ALL_SHARDS: self.get_primaries,
741 RequestPolicy.ALL_NODES: self.get_nodes,
742 RequestPolicy.ALL_REPLICAS: self.get_replicas,
743 RequestPolicy.MULTI_SHARD: lambda *args,
744 **kwargs: self._split_multi_shard_command(*args, **kwargs),
745 RequestPolicy.SPECIAL: self.get_special_nodes,
746 ResponsePolicy.DEFAULT_KEYLESS: lambda res: res,
747 ResponsePolicy.DEFAULT_KEYED: lambda res: res,
748 }
750 self._policy_resolver = policy_resolver
751 self.commands_parser = CommandsParser(self)
753 # Node where FT.AGGREGATE command is executed.
754 self._aggregate_nodes = None
755 self._lock = threading.RLock()
757 def __enter__(self):
758 return self
760 def __exit__(self, exc_type, exc_value, traceback):
761 self.close()
763 def __del__(self):
764 try:
765 self.close()
766 except Exception:
767 pass
769 def disconnect_connection_pools(self):
770 for node in self.get_nodes():
771 if node.redis_connection:
772 try:
773 node.redis_connection.connection_pool.disconnect()
774 except OSError:
775 # Client was already disconnected. do nothing
776 pass
778 def on_connect(self, connection):
779 """
780 Initialize the connection, authenticate and select a database and send
781 READONLY if it is set during object initialization.
782 """
783 connection.on_connect()
785 if self.read_from_replicas or self.load_balancing_strategy:
786 # Sending READONLY command to server to configure connection as
787 # readonly. Since each cluster node may change its server type due
788 # to a failover, we should establish a READONLY connection
789 # regardless of the server type. If this is a primary connection,
790 # READONLY would not affect executing write commands.
791 connection.send_command("READONLY")
792 if str_if_bytes(connection.read_response()) != "OK":
793 raise ConnectionError("READONLY command failed")
795 if self.user_on_connect_func is not None:
796 self.user_on_connect_func(connection)
798 def get_redis_connection(self, node: "ClusterNode") -> Redis:
799 if not node.redis_connection:
800 with self._lock:
801 if not node.redis_connection:
802 self.nodes_manager.create_redis_connections([node])
803 return node.redis_connection
805 def get_node(self, host=None, port=None, node_name=None):
806 return self.nodes_manager.get_node(host, port, node_name)
808 def get_primaries(self):
809 return self.nodes_manager.get_nodes_by_server_type(PRIMARY)
811 def get_replicas(self):
812 return self.nodes_manager.get_nodes_by_server_type(REPLICA)
814 def get_random_node(self):
815 return random.choice(list(self.nodes_manager.nodes_cache.values()))
817 def get_random_primary_or_all_nodes(self, command_name):
818 """
819 Returns random primary or all nodes depends on READONLY mode.
820 """
821 if self.read_from_replicas and command_name in READ_COMMANDS:
822 return self.get_random_node()
824 return self.get_random_primary_node()
826 def get_nodes(self):
827 return list(self.nodes_manager.nodes_cache.values())
829 def get_node_from_key(self, key, replica=False):
830 """
831 Get the node that holds the key's slot.
832 If replica set to True but the slot doesn't have any replicas, None is
833 returned.
834 """
835 slot = self.keyslot(key)
836 slot_cache = self.nodes_manager.slots_cache.get(slot)
837 if slot_cache is None or len(slot_cache) == 0:
838 raise SlotNotCoveredError(f'Slot "{slot}" is not covered by the cluster.')
839 if replica and len(self.nodes_manager.slots_cache[slot]) < 2:
840 return None
841 elif replica:
842 node_idx = 1
843 else:
844 # primary
845 node_idx = 0
847 return slot_cache[node_idx]
849 def get_default_node(self):
850 """
851 Get the cluster's default node
852 """
853 return self.nodes_manager.default_node
855 def get_nodes_from_slot(self, command: str, *args):
856 """
857 Returns a list of nodes that hold the specified keys' slots.
858 """
859 # get the node that holds the key's slot
860 slot = self.determine_slot(*args)
861 node = self.nodes_manager.get_node_from_slot(
862 slot,
863 self.read_from_replicas and command in READ_COMMANDS,
864 self.load_balancing_strategy if command in READ_COMMANDS else None,
865 )
866 return [node]
868 def _split_multi_shard_command(self, *args, **kwargs) -> list[dict]:
869 """
870 Splits the command with Multi-Shard policy, to the multiple commands
871 """
872 keys = self._get_command_keys(*args)
873 commands = []
875 for key in keys:
876 commands.append(
877 {
878 "args": (args[0], key),
879 "kwargs": kwargs,
880 }
881 )
883 return commands
885 def get_special_nodes(self) -> Optional[list["ClusterNode"]]:
886 """
887 Returns a list of nodes for commands with a special policy.
888 """
889 if not self._aggregate_nodes:
890 raise RedisClusterException(
891 "Cannot execute FT.CURSOR commands without FT.AGGREGATE"
892 )
894 return self._aggregate_nodes
896 def get_random_primary_node(self) -> "ClusterNode":
897 """
898 Returns a random primary node
899 """
900 return random.choice(self.get_primaries())
902 def _evaluate_all_succeeded(self, res):
903 """
904 Evaluate the result of a command with ResponsePolicy.ALL_SUCCEEDED
905 """
906 first_successful_response = None
908 if isinstance(res, dict):
909 for key, value in res.items():
910 if value:
911 if first_successful_response is None:
912 first_successful_response = {key: value}
913 else:
914 return {key: False}
915 else:
916 for response in res:
917 if response:
918 if first_successful_response is None:
919 # Dynamically resolve type
920 first_successful_response = type(response)(response)
921 else:
922 return type(response)(False)
924 return first_successful_response
926 def set_default_node(self, node):
927 """
928 Set the default node of the cluster.
929 :param node: 'ClusterNode'
930 :return True if the default node was set, else False
931 """
932 if node is None or self.get_node(node_name=node.name) is None:
933 return False
934 self.nodes_manager.default_node = node
935 return True
937 def set_retry(self, retry: Retry) -> None:
938 self.retry = retry
940 def monitor(self, target_node=None):
941 """
942 Returns a Monitor object for the specified target node.
943 The default cluster node will be selected if no target node was
944 specified.
945 Monitor is useful for handling the MONITOR command to the redis server.
946 next_command() method returns one command from monitor
947 listen() method yields commands from monitor.
948 """
949 if target_node is None:
950 target_node = self.get_default_node()
951 if target_node.redis_connection is None:
952 raise RedisClusterException(
953 f"Cluster Node {target_node.name} has no redis_connection"
954 )
955 return target_node.redis_connection.monitor()
957 def pubsub(self, node=None, host=None, port=None, **kwargs):
958 """
959 Allows passing a ClusterNode, or host&port, to get a pubsub instance
960 connected to the specified node
961 """
962 return ClusterPubSub(self, node=node, host=host, port=port, **kwargs)
964 def pipeline(self, transaction=None, shard_hint=None):
965 """
966 Cluster impl:
967 Pipelines do not work in cluster mode the same way they
968 do in normal mode. Create a clone of this object so
969 that simulating pipelines will work correctly. Each
970 command will be called directly when used and
971 when calling execute() will only return the result stack.
972 """
973 if shard_hint:
974 raise RedisClusterException("shard_hint is deprecated in cluster mode")
976 return ClusterPipeline(
977 nodes_manager=self.nodes_manager,
978 commands_parser=self.commands_parser,
979 startup_nodes=self.nodes_manager.startup_nodes,
980 result_callbacks=self.result_callbacks,
981 cluster_response_callbacks=self.cluster_response_callbacks,
982 read_from_replicas=self.read_from_replicas,
983 load_balancing_strategy=self.load_balancing_strategy,
984 reinitialize_steps=self.reinitialize_steps,
985 retry=self.retry,
986 lock=self._lock,
987 transaction=transaction,
988 )
990 def lock(
991 self,
992 name,
993 timeout=None,
994 sleep=0.1,
995 blocking=True,
996 blocking_timeout=None,
997 lock_class=None,
998 thread_local=True,
999 raise_on_release_error: bool = True,
1000 ):
1001 """
1002 Return a new Lock object using key ``name`` that mimics
1003 the behavior of threading.Lock.
1005 If specified, ``timeout`` indicates a maximum life for the lock.
1006 By default, it will remain locked until release() is called.
1008 ``sleep`` indicates the amount of time to sleep per loop iteration
1009 when the lock is in blocking mode and another client is currently
1010 holding the lock.
1012 ``blocking`` indicates whether calling ``acquire`` should block until
1013 the lock has been acquired or to fail immediately, causing ``acquire``
1014 to return False and the lock not being acquired. Defaults to True.
1015 Note this value can be overridden by passing a ``blocking``
1016 argument to ``acquire``.
1018 ``blocking_timeout`` indicates the maximum amount of time in seconds to
1019 spend trying to acquire the lock. A value of ``None`` indicates
1020 continue trying forever. ``blocking_timeout`` can be specified as a
1021 float or integer, both representing the number of seconds to wait.
1023 ``lock_class`` forces the specified lock implementation. Note that as
1024 of redis-py 3.0, the only lock class we implement is ``Lock`` (which is
1025 a Lua-based lock). So, it's unlikely you'll need this parameter, unless
1026 you have created your own custom lock class.
1028 ``thread_local`` indicates whether the lock token is placed in
1029 thread-local storage. By default, the token is placed in thread local
1030 storage so that a thread only sees its token, not a token set by
1031 another thread. Consider the following timeline:
1033 time: 0, thread-1 acquires `my-lock`, with a timeout of 5 seconds.
1034 thread-1 sets the token to "abc"
1035 time: 1, thread-2 blocks trying to acquire `my-lock` using the
1036 Lock instance.
1037 time: 5, thread-1 has not yet completed. redis expires the lock
1038 key.
1039 time: 5, thread-2 acquired `my-lock` now that it's available.
1040 thread-2 sets the token to "xyz"
1041 time: 6, thread-1 finishes its work and calls release(). if the
1042 token is *not* stored in thread local storage, then
1043 thread-1 would see the token value as "xyz" and would be
1044 able to successfully release the thread-2's lock.
1046 ``raise_on_release_error`` indicates whether to raise an exception when
1047 the lock is no longer owned when exiting the context manager. By default,
1048 this is True, meaning an exception will be raised. If False, the warning
1049 will be logged and the exception will be suppressed.
1051 In some use cases it's necessary to disable thread local storage. For
1052 example, if you have code where one thread acquires a lock and passes
1053 that lock instance to a worker thread to release later. If thread
1054 local storage isn't disabled in this case, the worker thread won't see
1055 the token set by the thread that acquired the lock. Our assumption
1056 is that these cases aren't common and as such default to using
1057 thread local storage."""
1058 if lock_class is None:
1059 lock_class = Lock
1060 return lock_class(
1061 self,
1062 name,
1063 timeout=timeout,
1064 sleep=sleep,
1065 blocking=blocking,
1066 blocking_timeout=blocking_timeout,
1067 thread_local=thread_local,
1068 raise_on_release_error=raise_on_release_error,
1069 )
1071 def set_response_callback(self, command, callback):
1072 """Set a custom Response Callback"""
1073 self.cluster_response_callbacks[command] = callback
1075 def _determine_nodes(
1076 self, *args, request_policy: RequestPolicy, **kwargs
1077 ) -> List["ClusterNode"]:
1078 """
1079 Determines a nodes the command should be executed on.
1080 """
1081 command = args[0].upper()
1082 if len(args) >= 2 and f"{args[0]} {args[1]}".upper() in self.command_flags:
1083 command = f"{args[0]} {args[1]}".upper()
1085 nodes_flag = kwargs.pop("nodes_flag", None)
1086 if nodes_flag is not None:
1087 # nodes flag passed by the user
1088 command_flag = nodes_flag
1089 else:
1090 # get the nodes group for this command if it was predefined
1091 command_flag = self.command_flags.get(command)
1093 if command_flag in self._command_flags_mapping:
1094 request_policy = self._command_flags_mapping[command_flag]
1096 policy_callback = self._policies_callback_mapping[request_policy]
1098 if request_policy == RequestPolicy.DEFAULT_KEYED:
1099 nodes = policy_callback(command, *args)
1100 elif request_policy == RequestPolicy.MULTI_SHARD:
1101 nodes = policy_callback(*args, **kwargs)
1102 elif request_policy == RequestPolicy.DEFAULT_KEYLESS:
1103 nodes = policy_callback(args[0])
1104 else:
1105 nodes = policy_callback()
1107 if args[0].lower() == "ft.aggregate":
1108 self._aggregate_nodes = nodes
1110 return nodes
1112 def _should_reinitialized(self):
1113 # To reinitialize the cluster on every MOVED error,
1114 # set reinitialize_steps to 1.
1115 # To avoid reinitializing the cluster on moved errors, set
1116 # reinitialize_steps to 0.
1117 if self.reinitialize_steps == 0:
1118 return False
1119 else:
1120 return self.reinitialize_counter % self.reinitialize_steps == 0
1122 def keyslot(self, key):
1123 """
1124 Calculate keyslot for a given key.
1125 See Keys distribution model in https://redis.io/topics/cluster-spec
1126 """
1127 k = self.encoder.encode(key)
1128 return key_slot(k)
1130 def _get_command_keys(self, *args):
1131 """
1132 Get the keys in the command. If the command has no keys in in, None is
1133 returned.
1135 NOTE: Due to a bug in redis<7.0, this function does not work properly
1136 for EVAL or EVALSHA when the `numkeys` arg is 0.
1137 - issue: https://github.com/redis/redis/issues/9493
1138 - fix: https://github.com/redis/redis/pull/9733
1140 So, don't use this function with EVAL or EVALSHA.
1141 """
1142 redis_conn = self.get_default_node().redis_connection
1143 return self.commands_parser.get_keys(redis_conn, *args)
1145 def determine_slot(self, *args) -> Optional[int]:
1146 """
1147 Figure out what slot to use based on args.
1149 Raises a RedisClusterException if there's a missing key and we can't
1150 determine what slots to map the command to; or, if the keys don't
1151 all map to the same key slot.
1152 """
1153 command = args[0]
1154 if self.command_flags.get(command) == SLOT_ID:
1155 # The command contains the slot ID
1156 return args[1]
1158 # Get the keys in the command
1160 # CLIENT TRACKING is a special case.
1161 # It doesn't have any keys, it needs to be sent to the provided nodes
1162 # By default it will be sent to all nodes.
1163 if command.upper() == "CLIENT TRACKING":
1164 return None
1166 # EVAL and EVALSHA are common enough that it's wasteful to go to the
1167 # redis server to parse the keys. Besides, there is a bug in redis<7.0
1168 # where `self._get_command_keys()` fails anyway. So, we special case
1169 # EVAL/EVALSHA.
1170 if command.upper() in ("EVAL", "EVALSHA"):
1171 # command syntax: EVAL "script body" num_keys ...
1172 if len(args) <= 2:
1173 raise RedisClusterException(f"Invalid args in command: {args}")
1174 num_actual_keys = int(args[2])
1175 eval_keys = args[3 : 3 + num_actual_keys]
1176 # if there are 0 keys, that means the script can be run on any node
1177 # so we can just return a random slot
1178 if len(eval_keys) == 0:
1179 return random.randrange(0, REDIS_CLUSTER_HASH_SLOTS)
1180 keys = eval_keys
1181 else:
1182 keys = self._get_command_keys(*args)
1183 if keys is None or len(keys) == 0:
1184 # FCALL can call a function with 0 keys, that means the function
1185 # can be run on any node so we can just return a random slot
1186 if command.upper() in ("FCALL", "FCALL_RO"):
1187 return random.randrange(0, REDIS_CLUSTER_HASH_SLOTS)
1188 raise RedisClusterException(
1189 "No way to dispatch this command to Redis Cluster. "
1190 "Missing key.\nYou can execute the command by specifying "
1191 f"target nodes.\nCommand: {args}"
1192 )
1194 # single key command
1195 if len(keys) == 1:
1196 return self.keyslot(keys[0])
1198 # multi-key command; we need to make sure all keys are mapped to
1199 # the same slot
1200 slots = {self.keyslot(key) for key in keys}
1201 if len(slots) != 1:
1202 raise RedisClusterException(
1203 f"{command} - all keys must map to the same key slot"
1204 )
1206 return slots.pop()
1208 def get_encoder(self):
1209 """
1210 Get the connections' encoder
1211 """
1212 return self.encoder
1214 def get_connection_kwargs(self):
1215 """
1216 Get the connections' key-word arguments
1217 """
1218 return self.nodes_manager.connection_kwargs
1220 def _is_nodes_flag(self, target_nodes):
1221 return isinstance(target_nodes, str) and target_nodes in self.node_flags
1223 def _parse_target_nodes(self, target_nodes):
1224 if isinstance(target_nodes, list):
1225 nodes = target_nodes
1226 elif isinstance(target_nodes, ClusterNode):
1227 # Supports passing a single ClusterNode as a variable
1228 nodes = [target_nodes]
1229 elif isinstance(target_nodes, dict):
1230 # Supports dictionaries of the format {node_name: node}.
1231 # It enables to execute commands with multi nodes as follows:
1232 # rc.cluster_save_config(rc.get_primaries())
1233 nodes = target_nodes.values()
1234 else:
1235 raise TypeError(
1236 "target_nodes type can be one of the following: "
1237 "node_flag (PRIMARIES, REPLICAS, RANDOM, ALL_NODES),"
1238 "ClusterNode, list<ClusterNode>, or dict<any, ClusterNode>. "
1239 f"The passed type is {type(target_nodes)}"
1240 )
1241 return nodes
1243 def execute_command(self, *args, **kwargs):
1244 return self._internal_execute_command(*args, **kwargs)
1246 def _internal_execute_command(self, *args, **kwargs):
1247 """
1248 Wrapper for ERRORS_ALLOW_RETRY error handling.
1250 It will try the number of times specified by the retries property from
1251 config option "self.retry" which defaults to 3 unless manually
1252 configured.
1254 If it reaches the number of times, the command will raise the exception
1256 Key argument :target_nodes: can be passed with the following types:
1257 nodes_flag: PRIMARIES, REPLICAS, ALL_NODES, RANDOM
1258 ClusterNode
1259 list<ClusterNode>
1260 dict<Any, ClusterNode>
1261 """
1262 target_nodes_specified = False
1263 is_default_node = False
1264 target_nodes = None
1265 passed_targets = kwargs.pop("target_nodes", None)
1266 command_policies = self._policy_resolver.resolve(args[0].lower())
1268 if passed_targets is not None and not self._is_nodes_flag(passed_targets):
1269 target_nodes = self._parse_target_nodes(passed_targets)
1270 target_nodes_specified = True
1272 if not command_policies and not target_nodes_specified:
1273 command = args[0].upper()
1274 if len(args) >= 2 and f"{args[0]} {args[1]}".upper() in self.command_flags:
1275 command = f"{args[0]} {args[1]}".upper()
1277 # We only could resolve key properties if command is not
1278 # in a list of pre-defined request policies
1279 command_flag = self.command_flags.get(command)
1280 if not command_flag:
1281 # Fallback to default policy
1282 if not self.get_default_node():
1283 slot = None
1284 else:
1285 slot = self.determine_slot(*args)
1286 if not slot:
1287 command_policies = CommandPolicies()
1288 else:
1289 command_policies = CommandPolicies(
1290 request_policy=RequestPolicy.DEFAULT_KEYED,
1291 response_policy=ResponsePolicy.DEFAULT_KEYED,
1292 )
1293 else:
1294 if command_flag in self._command_flags_mapping:
1295 command_policies = CommandPolicies(
1296 request_policy=self._command_flags_mapping[command_flag]
1297 )
1298 else:
1299 command_policies = CommandPolicies()
1300 elif not command_policies and target_nodes_specified:
1301 command_policies = CommandPolicies()
1303 # If an error that allows retrying was thrown, the nodes and slots
1304 # cache were reinitialized. We will retry executing the command with
1305 # the updated cluster setup only when the target nodes can be
1306 # determined again with the new cache tables. Therefore, when target
1307 # nodes were passed to this function, we cannot retry the command
1308 # execution since the nodes may not be valid anymore after the tables
1309 # were reinitialized. So in case of passed target nodes,
1310 # retry_attempts will be set to 0.
1311 retry_attempts = 0 if target_nodes_specified else self.retry.get_retries()
1312 # Add one for the first execution
1313 execute_attempts = 1 + retry_attempts
1314 for _ in range(execute_attempts):
1315 try:
1316 res = {}
1317 if not target_nodes_specified:
1318 # Determine the nodes to execute the command on
1319 target_nodes = self._determine_nodes(
1320 *args,
1321 request_policy=command_policies.request_policy,
1322 nodes_flag=passed_targets,
1323 )
1324 if not target_nodes:
1325 raise RedisClusterException(
1326 f"No targets were found to execute {args} command on"
1327 )
1328 if (
1329 len(target_nodes) == 1
1330 and target_nodes[0] == self.get_default_node()
1331 ):
1332 is_default_node = True
1333 for node in target_nodes:
1334 res[node.name] = self._execute_command(node, *args, **kwargs)
1336 if command_policies.response_policy == ResponsePolicy.ONE_SUCCEEDED:
1337 break
1339 # Return the processed result
1340 return self._process_result(
1341 args[0],
1342 res,
1343 response_policy=command_policies.response_policy,
1344 **kwargs,
1345 )
1346 except Exception as e:
1347 if retry_attempts > 0 and type(e) in self.__class__.ERRORS_ALLOW_RETRY:
1348 if is_default_node:
1349 # Replace the default cluster node
1350 self.replace_default_node()
1351 # The nodes and slots cache were reinitialized.
1352 # Try again with the new cluster setup.
1353 retry_attempts -= 1
1354 continue
1355 else:
1356 # raise the exception
1357 raise e
1359 def _execute_command(self, target_node, *args, **kwargs):
1360 """
1361 Send a command to a node in the cluster
1362 """
1363 command = args[0]
1364 redis_node = None
1365 connection = None
1366 redirect_addr = None
1367 asking = False
1368 moved = False
1369 ttl = int(self.RedisClusterRequestTTL)
1371 while ttl > 0:
1372 ttl -= 1
1373 try:
1374 if asking:
1375 target_node = self.get_node(node_name=redirect_addr)
1376 elif moved:
1377 # MOVED occurred and the slots cache was updated,
1378 # refresh the target node
1379 slot = self.determine_slot(*args)
1380 target_node = self.nodes_manager.get_node_from_slot(
1381 slot,
1382 self.read_from_replicas and command in READ_COMMANDS,
1383 self.load_balancing_strategy
1384 if command in READ_COMMANDS
1385 else None,
1386 )
1387 moved = False
1389 redis_node = self.get_redis_connection(target_node)
1390 connection = get_connection(redis_node)
1391 if asking:
1392 connection.send_command("ASKING")
1393 redis_node.parse_response(connection, "ASKING", **kwargs)
1394 asking = False
1395 connection.send_command(*args, **kwargs)
1396 response = redis_node.parse_response(connection, command, **kwargs)
1398 # Remove keys entry, it needs only for cache.
1399 kwargs.pop("keys", None)
1401 if command in self.cluster_response_callbacks:
1402 response = self.cluster_response_callbacks[command](
1403 response, **kwargs
1404 )
1405 return response
1406 except AuthenticationError:
1407 raise
1408 except MaxConnectionsError:
1409 # MaxConnectionsError indicates client-side resource exhaustion
1410 # (too many connections in the pool), not a node failure.
1411 # Don't treat this as a node failure - just re-raise the error
1412 # without reinitializing the cluster.
1413 raise
1414 except (ConnectionError, TimeoutError) as e:
1415 # ConnectionError can also be raised if we couldn't get a
1416 # connection from the pool before timing out, so check that
1417 # this is an actual connection before attempting to disconnect.
1418 if connection is not None:
1419 connection.disconnect()
1421 # Remove the failed node from the startup nodes before we try
1422 # to reinitialize the cluster
1423 self.nodes_manager.startup_nodes.pop(target_node.name, None)
1424 # Reset the cluster node's connection
1425 target_node.redis_connection = None
1426 self.nodes_manager.initialize()
1427 raise e
1428 except MovedError as e:
1429 # First, we will try to patch the slots/nodes cache with the
1430 # redirected node output and try again. If MovedError exceeds
1431 # 'reinitialize_steps' number of times, we will force
1432 # reinitializing the tables, and then try again.
1433 # 'reinitialize_steps' counter will increase faster when
1434 # the same client object is shared between multiple threads. To
1435 # reduce the frequency you can set this variable in the
1436 # RedisCluster constructor.
1437 self.reinitialize_counter += 1
1438 if self._should_reinitialized():
1439 self.nodes_manager.initialize()
1440 # Reset the counter
1441 self.reinitialize_counter = 0
1442 else:
1443 self.nodes_manager.update_moved_exception(e)
1444 moved = True
1445 except TryAgainError:
1446 if ttl < self.RedisClusterRequestTTL / 2:
1447 time.sleep(0.05)
1448 except AskError as e:
1449 redirect_addr = get_node_name(host=e.host, port=e.port)
1450 asking = True
1451 except (ClusterDownError, SlotNotCoveredError):
1452 # ClusterDownError can occur during a failover and to get
1453 # self-healed, we will try to reinitialize the cluster layout
1454 # and retry executing the command
1456 # SlotNotCoveredError can occur when the cluster is not fully
1457 # initialized or can be temporary issue.
1458 # We will try to reinitialize the cluster topology
1459 # and retry executing the command
1461 time.sleep(0.25)
1462 self.nodes_manager.initialize()
1463 raise
1464 except ResponseError:
1465 raise
1466 except Exception as e:
1467 if connection:
1468 connection.disconnect()
1469 raise e
1470 finally:
1471 if connection is not None:
1472 redis_node.connection_pool.release(connection)
1474 raise ClusterError("TTL exhausted.")
1476 def close(self) -> None:
1477 try:
1478 with self._lock:
1479 if self.nodes_manager:
1480 self.nodes_manager.close()
1481 except AttributeError:
1482 # RedisCluster's __init__ can fail before nodes_manager is set
1483 pass
1485 def _process_result(self, command, res, response_policy: ResponsePolicy, **kwargs):
1486 """
1487 Process the result of the executed command.
1488 The function would return a dict or a single value.
1490 :type command: str
1491 :type res: dict
1493 `res` should be in the following format:
1494 Dict<node_name, command_result>
1495 """
1496 if command in self.result_callbacks:
1497 res = self.result_callbacks[command](command, res, **kwargs)
1498 elif len(res) == 1:
1499 # When we execute the command on a single node, we can
1500 # remove the dictionary and return a single response
1501 res = list(res.values())[0]
1503 return self._policies_callback_mapping[response_policy](res)
1505 def load_external_module(self, funcname, func):
1506 """
1507 This function can be used to add externally defined redis modules,
1508 and their namespaces to the redis client.
1510 ``funcname`` - A string containing the name of the function to create
1511 ``func`` - The function, being added to this class.
1512 """
1513 setattr(self, funcname, func)
1515 def transaction(self, func, *watches, **kwargs):
1516 """
1517 Convenience method for executing the callable `func` as a transaction
1518 while watching all keys specified in `watches`. The 'func' callable
1519 should expect a single argument which is a Pipeline object.
1520 """
1521 shard_hint = kwargs.pop("shard_hint", None)
1522 value_from_callable = kwargs.pop("value_from_callable", False)
1523 watch_delay = kwargs.pop("watch_delay", None)
1524 with self.pipeline(True, shard_hint) as pipe:
1525 while True:
1526 try:
1527 if watches:
1528 pipe.watch(*watches)
1529 func_value = func(pipe)
1530 exec_value = pipe.execute()
1531 return func_value if value_from_callable else exec_value
1532 except WatchError:
1533 if watch_delay is not None and watch_delay > 0:
1534 time.sleep(watch_delay)
1535 continue
1538class ClusterNode:
1539 def __init__(self, host, port, server_type=None, redis_connection=None):
1540 if host == "localhost":
1541 host = socket.gethostbyname(host)
1543 self.host = host
1544 self.port = port
1545 self.name = get_node_name(host, port)
1546 self.server_type = server_type
1547 self.redis_connection = redis_connection
1549 def __repr__(self):
1550 return (
1551 f"[host={self.host},"
1552 f"port={self.port},"
1553 f"name={self.name},"
1554 f"server_type={self.server_type},"
1555 f"redis_connection={self.redis_connection}]"
1556 )
1558 def __eq__(self, obj):
1559 return isinstance(obj, ClusterNode) and obj.name == self.name
1561 def __del__(self):
1562 try:
1563 if self.redis_connection is not None:
1564 self.redis_connection.close()
1565 except Exception:
1566 # Ignore errors when closing the connection
1567 pass
1570class LoadBalancingStrategy(Enum):
1571 ROUND_ROBIN = "round_robin"
1572 ROUND_ROBIN_REPLICAS = "round_robin_replicas"
1573 RANDOM_REPLICA = "random_replica"
1576class LoadBalancer:
1577 """
1578 Round-Robin Load Balancing
1579 """
1581 def __init__(self, start_index: int = 0) -> None:
1582 self.primary_to_idx = {}
1583 self.start_index = start_index
1585 def get_server_index(
1586 self,
1587 primary: str,
1588 list_size: int,
1589 load_balancing_strategy: LoadBalancingStrategy = LoadBalancingStrategy.ROUND_ROBIN,
1590 ) -> int:
1591 if load_balancing_strategy == LoadBalancingStrategy.RANDOM_REPLICA:
1592 return self._get_random_replica_index(list_size)
1593 else:
1594 return self._get_round_robin_index(
1595 primary,
1596 list_size,
1597 load_balancing_strategy == LoadBalancingStrategy.ROUND_ROBIN_REPLICAS,
1598 )
1600 def reset(self) -> None:
1601 self.primary_to_idx.clear()
1603 def _get_random_replica_index(self, list_size: int) -> int:
1604 return random.randint(1, list_size - 1)
1606 def _get_round_robin_index(
1607 self, primary: str, list_size: int, replicas_only: bool
1608 ) -> int:
1609 server_index = self.primary_to_idx.setdefault(primary, self.start_index)
1610 if replicas_only and server_index == 0:
1611 # skip the primary node index
1612 server_index = 1
1613 # Update the index for the next round
1614 self.primary_to_idx[primary] = (server_index + 1) % list_size
1615 return server_index
1618class NodesManager:
1619 def __init__(
1620 self,
1621 startup_nodes,
1622 from_url=False,
1623 require_full_coverage=False,
1624 lock=None,
1625 dynamic_startup_nodes=True,
1626 connection_pool_class=ConnectionPool,
1627 address_remap: Optional[Callable[[Tuple[str, int]], Tuple[str, int]]] = None,
1628 cache: Optional[CacheInterface] = None,
1629 cache_config: Optional[CacheConfig] = None,
1630 cache_factory: Optional[CacheFactoryInterface] = None,
1631 event_dispatcher: Optional[EventDispatcher] = None,
1632 **kwargs,
1633 ):
1634 self.nodes_cache: Dict[str, Redis] = {}
1635 self.slots_cache = {}
1636 self.startup_nodes = {}
1637 self.default_node = None
1638 self.populate_startup_nodes(startup_nodes)
1639 self.from_url = from_url
1640 self._require_full_coverage = require_full_coverage
1641 self._dynamic_startup_nodes = dynamic_startup_nodes
1642 self.connection_pool_class = connection_pool_class
1643 self.address_remap = address_remap
1644 self._cache = cache
1645 self._cache_config = cache_config
1646 self._cache_factory = cache_factory
1647 self._moved_exception = None
1648 self.connection_kwargs = kwargs
1649 self.read_load_balancer = LoadBalancer()
1650 if lock is None:
1651 lock = threading.RLock()
1652 self._lock = lock
1653 if event_dispatcher is None:
1654 self._event_dispatcher = EventDispatcher()
1655 else:
1656 self._event_dispatcher = event_dispatcher
1657 self._credential_provider = self.connection_kwargs.get(
1658 "credential_provider", None
1659 )
1660 self.initialize()
1662 def get_node(self, host=None, port=None, node_name=None):
1663 """
1664 Get the requested node from the cluster's nodes.
1665 nodes.
1666 :return: ClusterNode if the node exists, else None
1667 """
1668 if host and port:
1669 # the user passed host and port
1670 if host == "localhost":
1671 host = socket.gethostbyname(host)
1672 return self.nodes_cache.get(get_node_name(host=host, port=port))
1673 elif node_name:
1674 return self.nodes_cache.get(node_name)
1675 else:
1676 return None
1678 def update_moved_exception(self, exception):
1679 self._moved_exception = exception
1681 def _update_moved_slots(self):
1682 """
1683 Update the slot's node with the redirected one
1684 """
1685 e = self._moved_exception
1686 redirected_node = self.get_node(host=e.host, port=e.port)
1687 if redirected_node is not None:
1688 # The node already exists
1689 if redirected_node.server_type is not PRIMARY:
1690 # Update the node's server type
1691 redirected_node.server_type = PRIMARY
1692 else:
1693 # This is a new node, we will add it to the nodes cache
1694 redirected_node = ClusterNode(e.host, e.port, PRIMARY)
1695 self.nodes_cache[redirected_node.name] = redirected_node
1696 slot_nodes = self.slots_cache[e.slot_id]
1697 if redirected_node not in slot_nodes:
1698 # The new slot owner is a new server, or a server from a different
1699 # shard. We need to remove all current nodes from the slot's list
1700 # (including replications) and add just the new node.
1701 self.slots_cache[e.slot_id] = [redirected_node]
1702 elif redirected_node is not slot_nodes[0]:
1703 # The MOVED error resulted from a failover, and the new slot owner
1704 # had previously been a replica.
1705 old_primary = slot_nodes[0]
1706 # Update the old primary to be a replica and add it to the end of
1707 # the slot's node list
1708 old_primary.server_type = REPLICA
1709 slot_nodes.append(old_primary)
1710 # Remove the old replica, which is now a primary, from the slot's
1711 # node list
1712 slot_nodes.remove(redirected_node)
1713 # Override the old primary with the new one
1714 slot_nodes[0] = redirected_node
1715 if self.default_node == old_primary:
1716 # Update the default node with the new primary
1717 self.default_node = redirected_node
1718 # else: circular MOVED to current primary -> no-op
1720 # Reset moved_exception
1721 self._moved_exception = None
1723 @deprecated_args(
1724 args_to_warn=["server_type"],
1725 reason=(
1726 "In case you need select some load balancing strategy "
1727 "that will use replicas, please set it through 'load_balancing_strategy'"
1728 ),
1729 version="5.3.0",
1730 )
1731 def get_node_from_slot(
1732 self,
1733 slot,
1734 read_from_replicas=False,
1735 load_balancing_strategy=None,
1736 server_type=None,
1737 ) -> ClusterNode:
1738 """
1739 Gets a node that servers this hash slot
1740 """
1741 if self._moved_exception:
1742 with self._lock:
1743 if self._moved_exception:
1744 self._update_moved_slots()
1746 if self.slots_cache.get(slot) is None or len(self.slots_cache[slot]) == 0:
1747 raise SlotNotCoveredError(
1748 f'Slot "{slot}" not covered by the cluster. '
1749 f'"require_full_coverage={self._require_full_coverage}"'
1750 )
1752 if read_from_replicas is True and load_balancing_strategy is None:
1753 load_balancing_strategy = LoadBalancingStrategy.ROUND_ROBIN
1755 if len(self.slots_cache[slot]) > 1 and load_balancing_strategy:
1756 # get the server index using the strategy defined in load_balancing_strategy
1757 primary_name = self.slots_cache[slot][0].name
1758 node_idx = self.read_load_balancer.get_server_index(
1759 primary_name, len(self.slots_cache[slot]), load_balancing_strategy
1760 )
1761 elif (
1762 server_type is None
1763 or server_type == PRIMARY
1764 or len(self.slots_cache[slot]) == 1
1765 ):
1766 # return a primary
1767 node_idx = 0
1768 else:
1769 # return a replica
1770 # randomly choose one of the replicas
1771 node_idx = random.randint(1, len(self.slots_cache[slot]) - 1)
1773 return self.slots_cache[slot][node_idx]
1775 def get_nodes_by_server_type(self, server_type):
1776 """
1777 Get all nodes with the specified server type
1778 :param server_type: 'primary' or 'replica'
1779 :return: list of ClusterNode
1780 """
1781 return [
1782 node
1783 for node in self.nodes_cache.values()
1784 if node.server_type == server_type
1785 ]
1787 def populate_startup_nodes(self, nodes):
1788 """
1789 Populate all startup nodes and filters out any duplicates
1790 """
1791 for n in nodes:
1792 self.startup_nodes[n.name] = n
1794 def check_slots_coverage(self, slots_cache):
1795 # Validate if all slots are covered or if we should try next
1796 # startup node
1797 for i in range(0, REDIS_CLUSTER_HASH_SLOTS):
1798 if i not in slots_cache:
1799 return False
1800 return True
1802 def create_redis_connections(self, nodes):
1803 """
1804 This function will create a redis connection to all nodes in :nodes:
1805 """
1806 connection_pools = []
1807 for node in nodes:
1808 if node.redis_connection is None:
1809 node.redis_connection = self.create_redis_node(
1810 host=node.host, port=node.port, **self.connection_kwargs
1811 )
1812 connection_pools.append(node.redis_connection.connection_pool)
1814 self._event_dispatcher.dispatch(
1815 AfterPooledConnectionsInstantiationEvent(
1816 connection_pools, ClientType.SYNC, self._credential_provider
1817 )
1818 )
1820 def create_redis_node(self, host, port, **kwargs):
1821 # We are configuring the connection pool not to retry
1822 # connections on lower level clients to avoid retrying
1823 # connections to nodes that are not reachable
1824 # and to avoid blocking the connection pool.
1825 # The only error that will have some handling in the lower
1826 # level clients is ConnectionError which will trigger disconnection
1827 # of the socket.
1828 # The retries will be handled on cluster client level
1829 # where we will have proper handling of the cluster topology
1830 node_retry_config = Retry(
1831 backoff=NoBackoff(), retries=0, supported_errors=(ConnectionError,)
1832 )
1834 protocol = kwargs.get("protocol", None)
1835 if protocol in [3, "3"]:
1836 kwargs.update(
1837 {"maint_notifications_config": MaintNotificationsConfig(enabled=False)}
1838 )
1839 if self.from_url:
1840 # Create a redis node with a costumed connection pool
1841 kwargs.update({"host": host})
1842 kwargs.update({"port": port})
1843 kwargs.update({"cache": self._cache})
1844 kwargs.update({"retry": node_retry_config})
1845 r = Redis(connection_pool=self.connection_pool_class(**kwargs))
1846 else:
1847 r = Redis(
1848 host=host,
1849 port=port,
1850 cache=self._cache,
1851 retry=node_retry_config,
1852 **kwargs,
1853 )
1854 return r
1856 def _get_or_create_cluster_node(self, host, port, role, tmp_nodes_cache):
1857 node_name = get_node_name(host, port)
1858 # check if we already have this node in the tmp_nodes_cache
1859 target_node = tmp_nodes_cache.get(node_name)
1860 if target_node is None:
1861 # before creating a new cluster node, check if the cluster node already
1862 # exists in the current nodes cache and has a valid connection so we can
1863 # reuse it
1864 target_node = self.nodes_cache.get(node_name)
1865 if target_node is None or target_node.redis_connection is None:
1866 # create new cluster node for this cluster
1867 target_node = ClusterNode(host, port, role)
1868 if target_node.server_type != role:
1869 target_node.server_type = role
1870 # add this node to the nodes cache
1871 tmp_nodes_cache[target_node.name] = target_node
1873 return target_node
1875 def initialize(self):
1876 """
1877 Initializes the nodes cache, slots cache and redis connections.
1878 :startup_nodes:
1879 Responsible for discovering other nodes in the cluster
1880 """
1881 self.reset()
1882 tmp_nodes_cache = {}
1883 tmp_slots = {}
1884 disagreements = []
1885 startup_nodes_reachable = False
1886 fully_covered = False
1887 kwargs = self.connection_kwargs
1888 exception = None
1889 # Convert to tuple to prevent RuntimeError if self.startup_nodes
1890 # is modified during iteration
1891 for startup_node in tuple(self.startup_nodes.values()):
1892 try:
1893 if startup_node.redis_connection:
1894 r = startup_node.redis_connection
1895 else:
1896 # Create a new Redis connection
1897 r = self.create_redis_node(
1898 startup_node.host, startup_node.port, **kwargs
1899 )
1900 self.startup_nodes[startup_node.name].redis_connection = r
1901 # Make sure cluster mode is enabled on this node
1902 try:
1903 cluster_slots = str_if_bytes(r.execute_command("CLUSTER SLOTS"))
1904 r.connection_pool.disconnect()
1905 except ResponseError:
1906 raise RedisClusterException(
1907 "Cluster mode is not enabled on this node"
1908 )
1909 startup_nodes_reachable = True
1910 except Exception as e:
1911 # Try the next startup node.
1912 # The exception is saved and raised only if we have no more nodes.
1913 exception = e
1914 continue
1916 # CLUSTER SLOTS command results in the following output:
1917 # [[slot_section[from_slot,to_slot,master,replica1,...,replicaN]]]
1918 # where each node contains the following list: [IP, port, node_id]
1919 # Therefore, cluster_slots[0][2][0] will be the IP address of the
1920 # primary node of the first slot section.
1921 # If there's only one server in the cluster, its ``host`` is ''
1922 # Fix it to the host in startup_nodes
1923 if (
1924 len(cluster_slots) == 1
1925 and len(cluster_slots[0][2][0]) == 0
1926 and len(self.startup_nodes) == 1
1927 ):
1928 cluster_slots[0][2][0] = startup_node.host
1930 for slot in cluster_slots:
1931 primary_node = slot[2]
1932 host = str_if_bytes(primary_node[0])
1933 if host == "":
1934 host = startup_node.host
1935 port = int(primary_node[1])
1936 host, port = self.remap_host_port(host, port)
1938 nodes_for_slot = []
1940 target_node = self._get_or_create_cluster_node(
1941 host, port, PRIMARY, tmp_nodes_cache
1942 )
1943 nodes_for_slot.append(target_node)
1945 replica_nodes = slot[3:]
1946 for replica_node in replica_nodes:
1947 host = str_if_bytes(replica_node[0])
1948 port = int(replica_node[1])
1949 host, port = self.remap_host_port(host, port)
1950 target_replica_node = self._get_or_create_cluster_node(
1951 host, port, REPLICA, tmp_nodes_cache
1952 )
1953 nodes_for_slot.append(target_replica_node)
1955 for i in range(int(slot[0]), int(slot[1]) + 1):
1956 if i not in tmp_slots:
1957 tmp_slots[i] = nodes_for_slot
1958 else:
1959 # Validate that 2 nodes want to use the same slot cache
1960 # setup
1961 tmp_slot = tmp_slots[i][0]
1962 if tmp_slot.name != target_node.name:
1963 disagreements.append(
1964 f"{tmp_slot.name} vs {target_node.name} on slot: {i}"
1965 )
1967 if len(disagreements) > 5:
1968 raise RedisClusterException(
1969 f"startup_nodes could not agree on a valid "
1970 f"slots cache: {', '.join(disagreements)}"
1971 )
1973 fully_covered = self.check_slots_coverage(tmp_slots)
1974 if fully_covered:
1975 # Don't need to continue to the next startup node if all
1976 # slots are covered
1977 break
1979 if not startup_nodes_reachable:
1980 raise RedisClusterException(
1981 f"Redis Cluster cannot be connected. Please provide at least "
1982 f"one reachable node: {str(exception)}"
1983 ) from exception
1985 if self._cache is None and self._cache_config is not None:
1986 if self._cache_factory is None:
1987 self._cache = CacheFactory(self._cache_config).get_cache()
1988 else:
1989 self._cache = self._cache_factory.get_cache()
1991 # Create Redis connections to all nodes
1992 self.create_redis_connections(list(tmp_nodes_cache.values()))
1994 # Check if the slots are not fully covered
1995 if not fully_covered and self._require_full_coverage:
1996 # Despite the requirement that the slots be covered, there
1997 # isn't a full coverage
1998 raise RedisClusterException(
1999 f"All slots are not covered after query all startup_nodes. "
2000 f"{len(tmp_slots)} of {REDIS_CLUSTER_HASH_SLOTS} "
2001 f"covered..."
2002 )
2004 # Set the tmp variables to the real variables
2005 self.nodes_cache = tmp_nodes_cache
2006 self.slots_cache = tmp_slots
2007 # Set the default node
2008 self.default_node = self.get_nodes_by_server_type(PRIMARY)[0]
2009 if self._dynamic_startup_nodes:
2010 # Populate the startup nodes with all discovered nodes
2011 self.startup_nodes = tmp_nodes_cache
2012 # If initialize was called after a MovedError, clear it
2013 self._moved_exception = None
2015 def close(self) -> None:
2016 self.default_node = None
2017 for node in self.nodes_cache.values():
2018 if node.redis_connection:
2019 node.redis_connection.close()
2021 def reset(self):
2022 try:
2023 self.read_load_balancer.reset()
2024 except TypeError:
2025 # The read_load_balancer is None, do nothing
2026 pass
2028 def remap_host_port(self, host: str, port: int) -> Tuple[str, int]:
2029 """
2030 Remap the host and port returned from the cluster to a different
2031 internal value. Useful if the client is not connecting directly
2032 to the cluster.
2033 """
2034 if self.address_remap:
2035 return self.address_remap((host, port))
2036 return host, port
2038 def find_connection_owner(self, connection: Connection) -> Optional[Redis]:
2039 node_name = get_node_name(connection.host, connection.port)
2040 for node in tuple(self.nodes_cache.values()):
2041 if node.redis_connection:
2042 conn_args = node.redis_connection.connection_pool.connection_kwargs
2043 if node_name == get_node_name(
2044 conn_args.get("host"), conn_args.get("port")
2045 ):
2046 return node
2049class ClusterPubSub(PubSub):
2050 """
2051 Wrapper for PubSub class.
2053 IMPORTANT: before using ClusterPubSub, read about the known limitations
2054 with pubsub in Cluster mode and learn how to workaround them:
2055 https://redis-py-cluster.readthedocs.io/en/stable/pubsub.html
2056 """
2058 def __init__(
2059 self,
2060 redis_cluster,
2061 node=None,
2062 host=None,
2063 port=None,
2064 push_handler_func=None,
2065 event_dispatcher: Optional["EventDispatcher"] = None,
2066 **kwargs,
2067 ):
2068 """
2069 When a pubsub instance is created without specifying a node, a single
2070 node will be transparently chosen for the pubsub connection on the
2071 first command execution. The node will be determined by:
2072 1. Hashing the channel name in the request to find its keyslot
2073 2. Selecting a node that handles the keyslot: If read_from_replicas is
2074 set to true or load_balancing_strategy is set, a replica can be selected.
2076 :type redis_cluster: RedisCluster
2077 :type node: ClusterNode
2078 :type host: str
2079 :type port: int
2080 """
2081 self.node = None
2082 self.set_pubsub_node(redis_cluster, node, host, port)
2083 connection_pool = (
2084 None
2085 if self.node is None
2086 else redis_cluster.get_redis_connection(self.node).connection_pool
2087 )
2088 self.cluster = redis_cluster
2089 self.node_pubsub_mapping = {}
2090 self._pubsubs_generator = self._pubsubs_generator()
2091 if event_dispatcher is None:
2092 self._event_dispatcher = EventDispatcher()
2093 else:
2094 self._event_dispatcher = event_dispatcher
2095 super().__init__(
2096 connection_pool=connection_pool,
2097 encoder=redis_cluster.encoder,
2098 push_handler_func=push_handler_func,
2099 event_dispatcher=self._event_dispatcher,
2100 **kwargs,
2101 )
2103 def set_pubsub_node(self, cluster, node=None, host=None, port=None):
2104 """
2105 The pubsub node will be set according to the passed node, host and port
2106 When none of the node, host, or port are specified - the node is set
2107 to None and will be determined by the keyslot of the channel in the
2108 first command to be executed.
2109 RedisClusterException will be thrown if the passed node does not exist
2110 in the cluster.
2111 If host is passed without port, or vice versa, a DataError will be
2112 thrown.
2113 :type cluster: RedisCluster
2114 :type node: ClusterNode
2115 :type host: str
2116 :type port: int
2117 """
2118 if node is not None:
2119 # node is passed by the user
2120 self._raise_on_invalid_node(cluster, node, node.host, node.port)
2121 pubsub_node = node
2122 elif host is not None and port is not None:
2123 # host and port passed by the user
2124 node = cluster.get_node(host=host, port=port)
2125 self._raise_on_invalid_node(cluster, node, host, port)
2126 pubsub_node = node
2127 elif any([host, port]) is True:
2128 # only 'host' or 'port' passed
2129 raise DataError("Passing a host requires passing a port, and vice versa")
2130 else:
2131 # nothing passed by the user. set node to None
2132 pubsub_node = None
2134 self.node = pubsub_node
2136 def get_pubsub_node(self):
2137 """
2138 Get the node that is being used as the pubsub connection
2139 """
2140 return self.node
2142 def _raise_on_invalid_node(self, redis_cluster, node, host, port):
2143 """
2144 Raise a RedisClusterException if the node is None or doesn't exist in
2145 the cluster.
2146 """
2147 if node is None or redis_cluster.get_node(node_name=node.name) is None:
2148 raise RedisClusterException(
2149 f"Node {host}:{port} doesn't exist in the cluster"
2150 )
2152 def execute_command(self, *args):
2153 """
2154 Execute a subscribe/unsubscribe command.
2156 Taken code from redis-py and tweak to make it work within a cluster.
2157 """
2158 # NOTE: don't parse the response in this function -- it could pull a
2159 # legitimate message off the stack if the connection is already
2160 # subscribed to one or more channels
2162 if self.connection is None:
2163 if self.connection_pool is None:
2164 if len(args) > 1:
2165 # Hash the first channel and get one of the nodes holding
2166 # this slot
2167 channel = args[1]
2168 slot = self.cluster.keyslot(channel)
2169 node = self.cluster.nodes_manager.get_node_from_slot(
2170 slot,
2171 self.cluster.read_from_replicas,
2172 self.cluster.load_balancing_strategy,
2173 )
2174 else:
2175 # Get a random node
2176 node = self.cluster.get_random_node()
2177 self.node = node
2178 redis_connection = self.cluster.get_redis_connection(node)
2179 self.connection_pool = redis_connection.connection_pool
2180 self.connection = self.connection_pool.get_connection()
2181 # register a callback that re-subscribes to any channels we
2182 # were listening to when we were disconnected
2183 self.connection.register_connect_callback(self.on_connect)
2184 if self.push_handler_func is not None:
2185 self.connection._parser.set_pubsub_push_handler(self.push_handler_func)
2186 self._event_dispatcher.dispatch(
2187 AfterPubSubConnectionInstantiationEvent(
2188 self.connection, self.connection_pool, ClientType.SYNC, self._lock
2189 )
2190 )
2191 connection = self.connection
2192 self._execute(connection, connection.send_command, *args)
2194 def _get_node_pubsub(self, node):
2195 try:
2196 return self.node_pubsub_mapping[node.name]
2197 except KeyError:
2198 pubsub = node.redis_connection.pubsub(
2199 push_handler_func=self.push_handler_func
2200 )
2201 self.node_pubsub_mapping[node.name] = pubsub
2202 return pubsub
2204 def _sharded_message_generator(self):
2205 for _ in range(len(self.node_pubsub_mapping)):
2206 pubsub = next(self._pubsubs_generator)
2207 message = pubsub.get_message()
2208 if message is not None:
2209 return message
2210 return None
2212 def _pubsubs_generator(self):
2213 while True:
2214 current_nodes = list(self.node_pubsub_mapping.values())
2215 yield from current_nodes
2217 def get_sharded_message(
2218 self, ignore_subscribe_messages=False, timeout=0.0, target_node=None
2219 ):
2220 if target_node:
2221 message = self.node_pubsub_mapping[target_node.name].get_message(
2222 ignore_subscribe_messages=ignore_subscribe_messages, timeout=timeout
2223 )
2224 else:
2225 message = self._sharded_message_generator()
2226 if message is None:
2227 return None
2228 elif str_if_bytes(message["type"]) == "sunsubscribe":
2229 if message["channel"] in self.pending_unsubscribe_shard_channels:
2230 self.pending_unsubscribe_shard_channels.remove(message["channel"])
2231 self.shard_channels.pop(message["channel"], None)
2232 node = self.cluster.get_node_from_key(message["channel"])
2233 if self.node_pubsub_mapping[node.name].subscribed is False:
2234 self.node_pubsub_mapping.pop(node.name)
2235 if not self.channels and not self.patterns and not self.shard_channels:
2236 # There are no subscriptions anymore, set subscribed_event flag
2237 # to false
2238 self.subscribed_event.clear()
2239 if self.ignore_subscribe_messages or ignore_subscribe_messages:
2240 return None
2241 return message
2243 def ssubscribe(self, *args, **kwargs):
2244 if args:
2245 args = list_or_args(args[0], args[1:])
2246 s_channels = dict.fromkeys(args)
2247 s_channels.update(kwargs)
2248 for s_channel, handler in s_channels.items():
2249 node = self.cluster.get_node_from_key(s_channel)
2250 pubsub = self._get_node_pubsub(node)
2251 if handler:
2252 pubsub.ssubscribe(**{s_channel: handler})
2253 else:
2254 pubsub.ssubscribe(s_channel)
2255 self.shard_channels.update(pubsub.shard_channels)
2256 self.pending_unsubscribe_shard_channels.difference_update(
2257 self._normalize_keys({s_channel: None})
2258 )
2259 if pubsub.subscribed and not self.subscribed:
2260 self.subscribed_event.set()
2261 self.health_check_response_counter = 0
2263 def sunsubscribe(self, *args):
2264 if args:
2265 args = list_or_args(args[0], args[1:])
2266 else:
2267 args = self.shard_channels
2269 for s_channel in args:
2270 node = self.cluster.get_node_from_key(s_channel)
2271 p = self._get_node_pubsub(node)
2272 p.sunsubscribe(s_channel)
2273 self.pending_unsubscribe_shard_channels.update(
2274 p.pending_unsubscribe_shard_channels
2275 )
2277 def get_redis_connection(self):
2278 """
2279 Get the Redis connection of the pubsub connected node.
2280 """
2281 if self.node is not None:
2282 return self.node.redis_connection
2284 def disconnect(self):
2285 """
2286 Disconnect the pubsub connection.
2287 """
2288 if self.connection:
2289 self.connection.disconnect()
2290 for pubsub in self.node_pubsub_mapping.values():
2291 pubsub.connection.disconnect()
2294class ClusterPipeline(RedisCluster):
2295 """
2296 Support for Redis pipeline
2297 in cluster mode
2298 """
2300 ERRORS_ALLOW_RETRY = (
2301 ConnectionError,
2302 TimeoutError,
2303 MovedError,
2304 AskError,
2305 TryAgainError,
2306 )
2308 NO_SLOTS_COMMANDS = {"UNWATCH"}
2309 IMMEDIATE_EXECUTE_COMMANDS = {"WATCH", "UNWATCH"}
2310 UNWATCH_COMMANDS = {"DISCARD", "EXEC", "UNWATCH"}
2312 @deprecated_args(
2313 args_to_warn=[
2314 "cluster_error_retry_attempts",
2315 ],
2316 reason="Please configure the 'retry' object instead",
2317 version="6.0.0",
2318 )
2319 def __init__(
2320 self,
2321 nodes_manager: "NodesManager",
2322 commands_parser: "CommandsParser",
2323 result_callbacks: Optional[Dict[str, Callable]] = None,
2324 cluster_response_callbacks: Optional[Dict[str, Callable]] = None,
2325 startup_nodes: Optional[List["ClusterNode"]] = None,
2326 read_from_replicas: bool = False,
2327 load_balancing_strategy: Optional[LoadBalancingStrategy] = None,
2328 cluster_error_retry_attempts: int = 3,
2329 reinitialize_steps: int = 5,
2330 retry: Optional[Retry] = None,
2331 lock=None,
2332 transaction=False,
2333 policy_resolver: PolicyResolver = StaticPolicyResolver(),
2334 **kwargs,
2335 ):
2336 """ """
2337 self.command_stack = []
2338 self.nodes_manager = nodes_manager
2339 self.commands_parser = commands_parser
2340 self.refresh_table_asap = False
2341 self.result_callbacks = (
2342 result_callbacks or self.__class__.RESULT_CALLBACKS.copy()
2343 )
2344 self.startup_nodes = startup_nodes if startup_nodes else []
2345 self.read_from_replicas = read_from_replicas
2346 self.load_balancing_strategy = load_balancing_strategy
2347 self.command_flags = self.__class__.COMMAND_FLAGS.copy()
2348 self.cluster_response_callbacks = cluster_response_callbacks
2349 self.reinitialize_counter = 0
2350 self.reinitialize_steps = reinitialize_steps
2351 if retry is not None:
2352 self.retry = retry
2353 else:
2354 self.retry = Retry(
2355 backoff=ExponentialWithJitterBackoff(base=1, cap=10),
2356 retries=cluster_error_retry_attempts,
2357 )
2359 self.encoder = Encoder(
2360 kwargs.get("encoding", "utf-8"),
2361 kwargs.get("encoding_errors", "strict"),
2362 kwargs.get("decode_responses", False),
2363 )
2364 if lock is None:
2365 lock = threading.RLock()
2366 self._lock = lock
2367 self.parent_execute_command = super().execute_command
2368 self._execution_strategy: ExecutionStrategy = (
2369 PipelineStrategy(self) if not transaction else TransactionStrategy(self)
2370 )
2372 # For backward compatibility, mapping from existing policies to new one
2373 self._command_flags_mapping: dict[str, Union[RequestPolicy, ResponsePolicy]] = {
2374 self.__class__.RANDOM: RequestPolicy.DEFAULT_KEYLESS,
2375 self.__class__.PRIMARIES: RequestPolicy.ALL_SHARDS,
2376 self.__class__.ALL_NODES: RequestPolicy.ALL_NODES,
2377 self.__class__.REPLICAS: RequestPolicy.ALL_REPLICAS,
2378 self.__class__.DEFAULT_NODE: RequestPolicy.DEFAULT_NODE,
2379 SLOT_ID: RequestPolicy.DEFAULT_KEYED,
2380 }
2382 self._policies_callback_mapping: dict[
2383 Union[RequestPolicy, ResponsePolicy], Callable
2384 ] = {
2385 RequestPolicy.DEFAULT_KEYLESS: lambda command_name: [
2386 self.get_random_primary_or_all_nodes(command_name)
2387 ],
2388 RequestPolicy.DEFAULT_KEYED: lambda command,
2389 *args: self.get_nodes_from_slot(command, *args),
2390 RequestPolicy.DEFAULT_NODE: lambda: [self.get_default_node()],
2391 RequestPolicy.ALL_SHARDS: self.get_primaries,
2392 RequestPolicy.ALL_NODES: self.get_nodes,
2393 RequestPolicy.ALL_REPLICAS: self.get_replicas,
2394 RequestPolicy.MULTI_SHARD: lambda *args,
2395 **kwargs: self._split_multi_shard_command(*args, **kwargs),
2396 RequestPolicy.SPECIAL: self.get_special_nodes,
2397 ResponsePolicy.DEFAULT_KEYLESS: lambda res: res,
2398 ResponsePolicy.DEFAULT_KEYED: lambda res: res,
2399 }
2401 self._policy_resolver = policy_resolver
2403 def __repr__(self):
2404 """ """
2405 return f"{type(self).__name__}"
2407 def __enter__(self):
2408 """ """
2409 return self
2411 def __exit__(self, exc_type, exc_value, traceback):
2412 """ """
2413 self.reset()
2415 def __del__(self):
2416 try:
2417 self.reset()
2418 except Exception:
2419 pass
2421 def __len__(self):
2422 """ """
2423 return len(self._execution_strategy.command_queue)
2425 def __bool__(self):
2426 "Pipeline instances should always evaluate to True on Python 3+"
2427 return True
2429 def execute_command(self, *args, **kwargs):
2430 """
2431 Wrapper function for pipeline_execute_command
2432 """
2433 return self._execution_strategy.execute_command(*args, **kwargs)
2435 def pipeline_execute_command(self, *args, **options):
2436 """
2437 Stage a command to be executed when execute() is next called
2439 Returns the current Pipeline object back so commands can be
2440 chained together, such as:
2442 pipe = pipe.set('foo', 'bar').incr('baz').decr('bang')
2444 At some other point, you can then run: pipe.execute(),
2445 which will execute all commands queued in the pipe.
2446 """
2447 return self._execution_strategy.execute_command(*args, **options)
2449 def annotate_exception(self, exception, number, command):
2450 """
2451 Provides extra context to the exception prior to it being handled
2452 """
2453 self._execution_strategy.annotate_exception(exception, number, command)
2455 def execute(self, raise_on_error: bool = True) -> List[Any]:
2456 """
2457 Execute all the commands in the current pipeline
2458 """
2460 try:
2461 return self._execution_strategy.execute(raise_on_error)
2462 finally:
2463 self.reset()
2465 def reset(self):
2466 """
2467 Reset back to empty pipeline.
2468 """
2469 self._execution_strategy.reset()
2471 def send_cluster_commands(
2472 self, stack, raise_on_error=True, allow_redirections=True
2473 ):
2474 return self._execution_strategy.send_cluster_commands(
2475 stack, raise_on_error=raise_on_error, allow_redirections=allow_redirections
2476 )
2478 def exists(self, *keys):
2479 return self._execution_strategy.exists(*keys)
2481 def eval(self):
2482 """ """
2483 return self._execution_strategy.eval()
2485 def multi(self):
2486 """
2487 Start a transactional block of the pipeline after WATCH commands
2488 are issued. End the transactional block with `execute`.
2489 """
2490 self._execution_strategy.multi()
2492 def load_scripts(self):
2493 """ """
2494 self._execution_strategy.load_scripts()
2496 def discard(self):
2497 """ """
2498 self._execution_strategy.discard()
2500 def watch(self, *names):
2501 """Watches the values at keys ``names``"""
2502 self._execution_strategy.watch(*names)
2504 def unwatch(self):
2505 """Unwatches all previously specified keys"""
2506 self._execution_strategy.unwatch()
2508 def script_load_for_pipeline(self, *args, **kwargs):
2509 self._execution_strategy.script_load_for_pipeline(*args, **kwargs)
2511 def delete(self, *names):
2512 self._execution_strategy.delete(*names)
2514 def unlink(self, *names):
2515 self._execution_strategy.unlink(*names)
2518def block_pipeline_command(name: str) -> Callable[..., Any]:
2519 """
2520 Prints error because some pipelined commands should
2521 be blocked when running in cluster-mode
2522 """
2524 def inner(*args, **kwargs):
2525 raise RedisClusterException(
2526 f"ERROR: Calling pipelined function {name} is blocked "
2527 f"when running redis in cluster mode..."
2528 )
2530 return inner
2533# Blocked pipeline commands
2534PIPELINE_BLOCKED_COMMANDS = (
2535 "BGREWRITEAOF",
2536 "BGSAVE",
2537 "BITOP",
2538 "BRPOPLPUSH",
2539 "CLIENT GETNAME",
2540 "CLIENT KILL",
2541 "CLIENT LIST",
2542 "CLIENT SETNAME",
2543 "CLIENT",
2544 "CONFIG GET",
2545 "CONFIG RESETSTAT",
2546 "CONFIG REWRITE",
2547 "CONFIG SET",
2548 "CONFIG",
2549 "DBSIZE",
2550 "ECHO",
2551 "EVALSHA",
2552 "FLUSHALL",
2553 "FLUSHDB",
2554 "INFO",
2555 "KEYS",
2556 "LASTSAVE",
2557 "MGET",
2558 "MGET NONATOMIC",
2559 "MOVE",
2560 "MSET",
2561 "MSETEX",
2562 "MSET NONATOMIC",
2563 "MSETNX",
2564 "PFCOUNT",
2565 "PFMERGE",
2566 "PING",
2567 "PUBLISH",
2568 "RANDOMKEY",
2569 "READONLY",
2570 "READWRITE",
2571 "RENAME",
2572 "RENAMENX",
2573 "RPOPLPUSH",
2574 "SAVE",
2575 "SCAN",
2576 "SCRIPT EXISTS",
2577 "SCRIPT FLUSH",
2578 "SCRIPT KILL",
2579 "SCRIPT LOAD",
2580 "SCRIPT",
2581 "SDIFF",
2582 "SDIFFSTORE",
2583 "SENTINEL GET MASTER ADDR BY NAME",
2584 "SENTINEL MASTER",
2585 "SENTINEL MASTERS",
2586 "SENTINEL MONITOR",
2587 "SENTINEL REMOVE",
2588 "SENTINEL SENTINELS",
2589 "SENTINEL SET",
2590 "SENTINEL SLAVES",
2591 "SENTINEL",
2592 "SHUTDOWN",
2593 "SINTER",
2594 "SINTERSTORE",
2595 "SLAVEOF",
2596 "SLOWLOG GET",
2597 "SLOWLOG LEN",
2598 "SLOWLOG RESET",
2599 "SLOWLOG",
2600 "SMOVE",
2601 "SORT",
2602 "SUNION",
2603 "SUNIONSTORE",
2604 "TIME",
2605)
2606for command in PIPELINE_BLOCKED_COMMANDS:
2607 command = command.replace(" ", "_").lower()
2609 setattr(ClusterPipeline, command, block_pipeline_command(command))
2612class PipelineCommand:
2613 """ """
2615 def __init__(self, args, options=None, position=None):
2616 self.args = args
2617 if options is None:
2618 options = {}
2619 self.options = options
2620 self.position = position
2621 self.result = None
2622 self.node = None
2623 self.asking = False
2624 self.command_policies: Optional[CommandPolicies] = None
2627class NodeCommands:
2628 """ """
2630 def __init__(self, parse_response, connection_pool, connection):
2631 """ """
2632 self.parse_response = parse_response
2633 self.connection_pool = connection_pool
2634 self.connection = connection
2635 self.commands = []
2637 def append(self, c):
2638 """ """
2639 self.commands.append(c)
2641 def write(self):
2642 """
2643 Code borrowed from Redis so it can be fixed
2644 """
2645 connection = self.connection
2646 commands = self.commands
2648 # We are going to clobber the commands with the write, so go ahead
2649 # and ensure that nothing is sitting there from a previous run.
2650 for c in commands:
2651 c.result = None
2653 # build up all commands into a single request to increase network perf
2654 # send all the commands and catch connection and timeout errors.
2655 try:
2656 connection.send_packed_command(
2657 connection.pack_commands([c.args for c in commands])
2658 )
2659 except (ConnectionError, TimeoutError) as e:
2660 for c in commands:
2661 c.result = e
2663 def read(self):
2664 """ """
2665 connection = self.connection
2666 for c in self.commands:
2667 # if there is a result on this command,
2668 # it means we ran into an exception
2669 # like a connection error. Trying to parse
2670 # a response on a connection that
2671 # is no longer open will result in a
2672 # connection error raised by redis-py.
2673 # but redis-py doesn't check in parse_response
2674 # that the sock object is
2675 # still set and if you try to
2676 # read from a closed connection, it will
2677 # result in an AttributeError because
2678 # it will do a readline() call on None.
2679 # This can have all kinds of nasty side-effects.
2680 # Treating this case as a connection error
2681 # is fine because it will dump
2682 # the connection object back into the
2683 # pool and on the next write, it will
2684 # explicitly open the connection and all will be well.
2685 if c.result is None:
2686 try:
2687 c.result = self.parse_response(connection, c.args[0], **c.options)
2688 except (ConnectionError, TimeoutError) as e:
2689 for c in self.commands:
2690 c.result = e
2691 return
2692 except RedisError:
2693 c.result = sys.exc_info()[1]
2696class ExecutionStrategy(ABC):
2697 @property
2698 @abstractmethod
2699 def command_queue(self):
2700 pass
2702 @abstractmethod
2703 def execute_command(self, *args, **kwargs):
2704 """
2705 Execution flow for current execution strategy.
2707 See: ClusterPipeline.execute_command()
2708 """
2709 pass
2711 @abstractmethod
2712 def annotate_exception(self, exception, number, command):
2713 """
2714 Annotate exception according to current execution strategy.
2716 See: ClusterPipeline.annotate_exception()
2717 """
2718 pass
2720 @abstractmethod
2721 def pipeline_execute_command(self, *args, **options):
2722 """
2723 Pipeline execution flow for current execution strategy.
2725 See: ClusterPipeline.pipeline_execute_command()
2726 """
2727 pass
2729 @abstractmethod
2730 def execute(self, raise_on_error: bool = True) -> List[Any]:
2731 """
2732 Executes current execution strategy.
2734 See: ClusterPipeline.execute()
2735 """
2736 pass
2738 @abstractmethod
2739 def send_cluster_commands(
2740 self, stack, raise_on_error=True, allow_redirections=True
2741 ):
2742 """
2743 Sends commands according to current execution strategy.
2745 See: ClusterPipeline.send_cluster_commands()
2746 """
2747 pass
2749 @abstractmethod
2750 def reset(self):
2751 """
2752 Resets current execution strategy.
2754 See: ClusterPipeline.reset()
2755 """
2756 pass
2758 @abstractmethod
2759 def exists(self, *keys):
2760 pass
2762 @abstractmethod
2763 def eval(self):
2764 pass
2766 @abstractmethod
2767 def multi(self):
2768 """
2769 Starts transactional context.
2771 See: ClusterPipeline.multi()
2772 """
2773 pass
2775 @abstractmethod
2776 def load_scripts(self):
2777 pass
2779 @abstractmethod
2780 def watch(self, *names):
2781 pass
2783 @abstractmethod
2784 def unwatch(self):
2785 """
2786 Unwatches all previously specified keys
2788 See: ClusterPipeline.unwatch()
2789 """
2790 pass
2792 @abstractmethod
2793 def script_load_for_pipeline(self, *args, **kwargs):
2794 pass
2796 @abstractmethod
2797 def delete(self, *names):
2798 """
2799 "Delete a key specified by ``names``"
2801 See: ClusterPipeline.delete()
2802 """
2803 pass
2805 @abstractmethod
2806 def unlink(self, *names):
2807 """
2808 "Unlink a key specified by ``names``"
2810 See: ClusterPipeline.unlink()
2811 """
2812 pass
2814 @abstractmethod
2815 def discard(self):
2816 pass
2819class AbstractStrategy(ExecutionStrategy):
2820 def __init__(
2821 self,
2822 pipe: ClusterPipeline,
2823 ):
2824 self._command_queue: List[PipelineCommand] = []
2825 self._pipe = pipe
2826 self._nodes_manager = self._pipe.nodes_manager
2828 @property
2829 def command_queue(self):
2830 return self._command_queue
2832 @command_queue.setter
2833 def command_queue(self, queue: List[PipelineCommand]):
2834 self._command_queue = queue
2836 @abstractmethod
2837 def execute_command(self, *args, **kwargs):
2838 pass
2840 def pipeline_execute_command(self, *args, **options):
2841 self._command_queue.append(
2842 PipelineCommand(args, options, len(self._command_queue))
2843 )
2844 return self._pipe
2846 @abstractmethod
2847 def execute(self, raise_on_error: bool = True) -> List[Any]:
2848 pass
2850 @abstractmethod
2851 def send_cluster_commands(
2852 self, stack, raise_on_error=True, allow_redirections=True
2853 ):
2854 pass
2856 @abstractmethod
2857 def reset(self):
2858 pass
2860 def exists(self, *keys):
2861 return self.execute_command("EXISTS", *keys)
2863 def eval(self):
2864 """ """
2865 raise RedisClusterException("method eval() is not implemented")
2867 def load_scripts(self):
2868 """ """
2869 raise RedisClusterException("method load_scripts() is not implemented")
2871 def script_load_for_pipeline(self, *args, **kwargs):
2872 """ """
2873 raise RedisClusterException(
2874 "method script_load_for_pipeline() is not implemented"
2875 )
2877 def annotate_exception(self, exception, number, command):
2878 """
2879 Provides extra context to the exception prior to it being handled
2880 """
2881 cmd = " ".join(map(safe_str, command))
2882 msg = (
2883 f"Command # {number} ({truncate_text(cmd)}) of pipeline "
2884 f"caused error: {exception.args[0]}"
2885 )
2886 exception.args = (msg,) + exception.args[1:]
2889class PipelineStrategy(AbstractStrategy):
2890 def __init__(self, pipe: ClusterPipeline):
2891 super().__init__(pipe)
2892 self.command_flags = pipe.command_flags
2894 def execute_command(self, *args, **kwargs):
2895 return self.pipeline_execute_command(*args, **kwargs)
2897 def _raise_first_error(self, stack):
2898 """
2899 Raise the first exception on the stack
2900 """
2901 for c in stack:
2902 r = c.result
2903 if isinstance(r, Exception):
2904 self.annotate_exception(r, c.position + 1, c.args)
2905 raise r
2907 def execute(self, raise_on_error: bool = True) -> List[Any]:
2908 stack = self._command_queue
2909 if not stack:
2910 return []
2912 try:
2913 return self.send_cluster_commands(stack, raise_on_error)
2914 finally:
2915 self.reset()
2917 def reset(self):
2918 """
2919 Reset back to empty pipeline.
2920 """
2921 self._command_queue = []
2923 def send_cluster_commands(
2924 self, stack, raise_on_error=True, allow_redirections=True
2925 ):
2926 """
2927 Wrapper for RedisCluster.ERRORS_ALLOW_RETRY errors handling.
2929 If one of the retryable exceptions has been thrown we assume that:
2930 - connection_pool was disconnected
2931 - connection_pool was reset
2932 - refresh_table_asap set to True
2934 It will try the number of times specified by
2935 the retries in config option "self.retry"
2936 which defaults to 3 unless manually configured.
2938 If it reaches the number of times, the command will
2939 raises ClusterDownException.
2940 """
2941 if not stack:
2942 return []
2943 retry_attempts = self._pipe.retry.get_retries()
2944 while True:
2945 try:
2946 return self._send_cluster_commands(
2947 stack,
2948 raise_on_error=raise_on_error,
2949 allow_redirections=allow_redirections,
2950 )
2951 except RedisCluster.ERRORS_ALLOW_RETRY as e:
2952 if retry_attempts > 0:
2953 # Try again with the new cluster setup. All other errors
2954 # should be raised.
2955 retry_attempts -= 1
2956 pass
2957 else:
2958 raise e
2960 def _send_cluster_commands(
2961 self, stack, raise_on_error=True, allow_redirections=True
2962 ):
2963 """
2964 Send a bunch of cluster commands to the redis cluster.
2966 `allow_redirections` If the pipeline should follow
2967 `ASK` & `MOVED` responses automatically. If set
2968 to false it will raise RedisClusterException.
2969 """
2970 # the first time sending the commands we send all of
2971 # the commands that were queued up.
2972 # if we have to run through it again, we only retry
2973 # the commands that failed.
2974 attempt = sorted(stack, key=lambda x: x.position)
2975 is_default_node = False
2976 # build a list of node objects based on node names we need to
2977 nodes = {}
2979 # as we move through each command that still needs to be processed,
2980 # we figure out the slot number that command maps to, then from
2981 # the slot determine the node.
2982 for c in attempt:
2983 command_policies = self._pipe._policy_resolver.resolve(c.args[0].lower())
2985 while True:
2986 # refer to our internal node -> slot table that
2987 # tells us where a given command should route to.
2988 # (it might be possible we have a cached node that no longer
2989 # exists in the cluster, which is why we do this in a loop)
2990 passed_targets = c.options.pop("target_nodes", None)
2991 if passed_targets and not self._is_nodes_flag(passed_targets):
2992 target_nodes = self._parse_target_nodes(passed_targets)
2994 if not command_policies:
2995 command_policies = CommandPolicies()
2996 else:
2997 if not command_policies:
2998 command = c.args[0].upper()
2999 if (
3000 len(c.args) >= 2
3001 and f"{c.args[0]} {c.args[1]}".upper()
3002 in self._pipe.command_flags
3003 ):
3004 command = f"{c.args[0]} {c.args[1]}".upper()
3006 # We only could resolve key properties if command is not
3007 # in a list of pre-defined request policies
3008 command_flag = self.command_flags.get(command)
3009 if not command_flag:
3010 # Fallback to default policy
3011 if not self._pipe.get_default_node():
3012 keys = None
3013 else:
3014 keys = self._pipe._get_command_keys(*c.args)
3015 if not keys or len(keys) == 0:
3016 command_policies = CommandPolicies()
3017 else:
3018 command_policies = CommandPolicies(
3019 request_policy=RequestPolicy.DEFAULT_KEYED,
3020 response_policy=ResponsePolicy.DEFAULT_KEYED,
3021 )
3022 else:
3023 if command_flag in self._pipe._command_flags_mapping:
3024 command_policies = CommandPolicies(
3025 request_policy=self._pipe._command_flags_mapping[
3026 command_flag
3027 ]
3028 )
3029 else:
3030 command_policies = CommandPolicies()
3032 target_nodes = self._determine_nodes(
3033 *c.args,
3034 request_policy=command_policies.request_policy,
3035 node_flag=passed_targets,
3036 )
3037 if not target_nodes:
3038 raise RedisClusterException(
3039 f"No targets were found to execute {c.args} command on"
3040 )
3041 c.command_policies = command_policies
3042 if len(target_nodes) > 1:
3043 raise RedisClusterException(
3044 f"Too many targets for command {c.args}"
3045 )
3047 node = target_nodes[0]
3048 if node == self._pipe.get_default_node():
3049 is_default_node = True
3051 # now that we know the name of the node
3052 # ( it's just a string in the form of host:port )
3053 # we can build a list of commands for each node.
3054 node_name = node.name
3055 if node_name not in nodes:
3056 redis_node = self._pipe.get_redis_connection(node)
3057 try:
3058 connection = get_connection(redis_node)
3059 except (ConnectionError, TimeoutError):
3060 for n in nodes.values():
3061 n.connection_pool.release(n.connection)
3062 # Connection retries are being handled in the node's
3063 # Retry object. Reinitialize the node -> slot table.
3064 self._nodes_manager.initialize()
3065 if is_default_node:
3066 self._pipe.replace_default_node()
3067 raise
3068 nodes[node_name] = NodeCommands(
3069 redis_node.parse_response,
3070 redis_node.connection_pool,
3071 connection,
3072 )
3073 nodes[node_name].append(c)
3074 break
3076 # send the commands in sequence.
3077 # we write to all the open sockets for each node first,
3078 # before reading anything
3079 # this allows us to flush all the requests out across the
3080 # network
3081 # so that we can read them from different sockets as they come back.
3082 # we dont' multiplex on the sockets as they come available,
3083 # but that shouldn't make too much difference.
3084 try:
3085 node_commands = nodes.values()
3086 for n in node_commands:
3087 n.write()
3089 for n in node_commands:
3090 n.read()
3091 finally:
3092 # release all of the redis connections we allocated earlier
3093 # back into the connection pool.
3094 # we used to do this step as part of a try/finally block,
3095 # but it is really dangerous to
3096 # release connections back into the pool if for some
3097 # reason the socket has data still left in it
3098 # from a previous operation. The write and
3099 # read operations already have try/catch around them for
3100 # all known types of errors including connection
3101 # and socket level errors.
3102 # So if we hit an exception, something really bad
3103 # happened and putting any oF
3104 # these connections back into the pool is a very bad idea.
3105 # the socket might have unread buffer still sitting in it,
3106 # and then the next time we read from it we pass the
3107 # buffered result back from a previous command and
3108 # every single request after to that connection will always get
3109 # a mismatched result.
3110 for n in nodes.values():
3111 n.connection_pool.release(n.connection)
3113 # if the response isn't an exception it is a
3114 # valid response from the node
3115 # we're all done with that command, YAY!
3116 # if we have more commands to attempt, we've run into problems.
3117 # collect all the commands we are allowed to retry.
3118 # (MOVED, ASK, or connection errors or timeout errors)
3119 attempt = sorted(
3120 (
3121 c
3122 for c in attempt
3123 if isinstance(c.result, ClusterPipeline.ERRORS_ALLOW_RETRY)
3124 ),
3125 key=lambda x: x.position,
3126 )
3127 if attempt and allow_redirections:
3128 # RETRY MAGIC HAPPENS HERE!
3129 # send these remaining commands one at a time using `execute_command`
3130 # in the main client. This keeps our retry logic
3131 # in one place mostly,
3132 # and allows us to be more confident in correctness of behavior.
3133 # at this point any speed gains from pipelining have been lost
3134 # anyway, so we might as well make the best
3135 # attempt to get the correct behavior.
3136 #
3137 # The client command will handle retries for each
3138 # individual command sequentially as we pass each
3139 # one into `execute_command`. Any exceptions
3140 # that bubble out should only appear once all
3141 # retries have been exhausted.
3142 #
3143 # If a lot of commands have failed, we'll be setting the
3144 # flag to rebuild the slots table from scratch.
3145 # So MOVED errors should correct themselves fairly quickly.
3146 self._pipe.reinitialize_counter += 1
3147 if self._pipe._should_reinitialized():
3148 self._nodes_manager.initialize()
3149 if is_default_node:
3150 self._pipe.replace_default_node()
3151 for c in attempt:
3152 try:
3153 # send each command individually like we
3154 # do in the main client.
3155 c.result = self._pipe.parent_execute_command(*c.args, **c.options)
3156 except RedisError as e:
3157 c.result = e
3159 # turn the response back into a simple flat array that corresponds
3160 # to the sequence of commands issued in the stack in pipeline.execute()
3161 response = []
3162 for c in sorted(stack, key=lambda x: x.position):
3163 if c.args[0] in self._pipe.cluster_response_callbacks:
3164 # Remove keys entry, it needs only for cache.
3165 c.options.pop("keys", None)
3166 c.result = self._pipe._policies_callback_mapping[
3167 c.command_policies.response_policy
3168 ](
3169 self._pipe.cluster_response_callbacks[c.args[0]](
3170 c.result, **c.options
3171 )
3172 )
3173 response.append(c.result)
3175 if raise_on_error:
3176 self._raise_first_error(stack)
3178 return response
3180 def _is_nodes_flag(self, target_nodes):
3181 return isinstance(target_nodes, str) and target_nodes in self._pipe.node_flags
3183 def _parse_target_nodes(self, target_nodes):
3184 if isinstance(target_nodes, list):
3185 nodes = target_nodes
3186 elif isinstance(target_nodes, ClusterNode):
3187 # Supports passing a single ClusterNode as a variable
3188 nodes = [target_nodes]
3189 elif isinstance(target_nodes, dict):
3190 # Supports dictionaries of the format {node_name: node}.
3191 # It enables to execute commands with multi nodes as follows:
3192 # rc.cluster_save_config(rc.get_primaries())
3193 nodes = target_nodes.values()
3194 else:
3195 raise TypeError(
3196 "target_nodes type can be one of the following: "
3197 "node_flag (PRIMARIES, REPLICAS, RANDOM, ALL_NODES),"
3198 "ClusterNode, list<ClusterNode>, or dict<any, ClusterNode>. "
3199 f"The passed type is {type(target_nodes)}"
3200 )
3201 return nodes
3203 def _determine_nodes(
3204 self, *args, request_policy: RequestPolicy, **kwargs
3205 ) -> List["ClusterNode"]:
3206 # Determine which nodes should be executed the command on.
3207 # Returns a list of target nodes.
3208 command = args[0].upper()
3209 if (
3210 len(args) >= 2
3211 and f"{args[0]} {args[1]}".upper() in self._pipe.command_flags
3212 ):
3213 command = f"{args[0]} {args[1]}".upper()
3215 nodes_flag = kwargs.pop("nodes_flag", None)
3216 if nodes_flag is not None:
3217 # nodes flag passed by the user
3218 command_flag = nodes_flag
3219 else:
3220 # get the nodes group for this command if it was predefined
3221 command_flag = self._pipe.command_flags.get(command)
3223 if command_flag in self._pipe._command_flags_mapping:
3224 request_policy = self._pipe._command_flags_mapping[command_flag]
3226 policy_callback = self._pipe._policies_callback_mapping[request_policy]
3228 if request_policy == RequestPolicy.DEFAULT_KEYED:
3229 nodes = policy_callback(command, *args)
3230 elif request_policy == RequestPolicy.MULTI_SHARD:
3231 nodes = policy_callback(*args, **kwargs)
3232 elif request_policy == RequestPolicy.DEFAULT_KEYLESS:
3233 nodes = policy_callback(args[0])
3234 else:
3235 nodes = policy_callback()
3237 if args[0].lower() == "ft.aggregate":
3238 self._aggregate_nodes = nodes
3240 return nodes
3242 def multi(self):
3243 raise RedisClusterException(
3244 "method multi() is not supported outside of transactional context"
3245 )
3247 def discard(self):
3248 raise RedisClusterException(
3249 "method discard() is not supported outside of transactional context"
3250 )
3252 def watch(self, *names):
3253 raise RedisClusterException(
3254 "method watch() is not supported outside of transactional context"
3255 )
3257 def unwatch(self, *names):
3258 raise RedisClusterException(
3259 "method unwatch() is not supported outside of transactional context"
3260 )
3262 def delete(self, *names):
3263 if len(names) != 1:
3264 raise RedisClusterException(
3265 "deleting multiple keys is not implemented in pipeline command"
3266 )
3268 return self.execute_command("DEL", names[0])
3270 def unlink(self, *names):
3271 if len(names) != 1:
3272 raise RedisClusterException(
3273 "unlinking multiple keys is not implemented in pipeline command"
3274 )
3276 return self.execute_command("UNLINK", names[0])
3279class TransactionStrategy(AbstractStrategy):
3280 NO_SLOTS_COMMANDS = {"UNWATCH"}
3281 IMMEDIATE_EXECUTE_COMMANDS = {"WATCH", "UNWATCH"}
3282 UNWATCH_COMMANDS = {"DISCARD", "EXEC", "UNWATCH"}
3283 SLOT_REDIRECT_ERRORS = (AskError, MovedError)
3284 CONNECTION_ERRORS = (
3285 ConnectionError,
3286 OSError,
3287 ClusterDownError,
3288 SlotNotCoveredError,
3289 )
3291 def __init__(self, pipe: ClusterPipeline):
3292 super().__init__(pipe)
3293 self._explicit_transaction = False
3294 self._watching = False
3295 self._pipeline_slots: Set[int] = set()
3296 self._transaction_connection: Optional[Connection] = None
3297 self._executing = False
3298 self._retry = copy(self._pipe.retry)
3299 self._retry.update_supported_errors(
3300 RedisCluster.ERRORS_ALLOW_RETRY + self.SLOT_REDIRECT_ERRORS
3301 )
3303 def _get_client_and_connection_for_transaction(self) -> Tuple[Redis, Connection]:
3304 """
3305 Find a connection for a pipeline transaction.
3307 For running an atomic transaction, watch keys ensure that contents have not been
3308 altered as long as the watch commands for those keys were sent over the same
3309 connection. So once we start watching a key, we fetch a connection to the
3310 node that owns that slot and reuse it.
3311 """
3312 if not self._pipeline_slots:
3313 raise RedisClusterException(
3314 "At least a command with a key is needed to identify a node"
3315 )
3317 node: ClusterNode = self._nodes_manager.get_node_from_slot(
3318 list(self._pipeline_slots)[0], False
3319 )
3320 redis_node: Redis = self._pipe.get_redis_connection(node)
3321 if self._transaction_connection:
3322 if not redis_node.connection_pool.owns_connection(
3323 self._transaction_connection
3324 ):
3325 previous_node = self._nodes_manager.find_connection_owner(
3326 self._transaction_connection
3327 )
3328 previous_node.connection_pool.release(self._transaction_connection)
3329 self._transaction_connection = None
3331 if not self._transaction_connection:
3332 self._transaction_connection = get_connection(redis_node)
3334 return redis_node, self._transaction_connection
3336 def execute_command(self, *args, **kwargs):
3337 slot_number: Optional[int] = None
3338 if args[0] not in ClusterPipeline.NO_SLOTS_COMMANDS:
3339 slot_number = self._pipe.determine_slot(*args)
3341 if (
3342 self._watching or args[0] in self.IMMEDIATE_EXECUTE_COMMANDS
3343 ) and not self._explicit_transaction:
3344 if args[0] == "WATCH":
3345 self._validate_watch()
3347 if slot_number is not None:
3348 if self._pipeline_slots and slot_number not in self._pipeline_slots:
3349 raise CrossSlotTransactionError(
3350 "Cannot watch or send commands on different slots"
3351 )
3353 self._pipeline_slots.add(slot_number)
3354 elif args[0] not in self.NO_SLOTS_COMMANDS:
3355 raise RedisClusterException(
3356 f"Cannot identify slot number for command: {args[0]},"
3357 "it cannot be triggered in a transaction"
3358 )
3360 return self._immediate_execute_command(*args, **kwargs)
3361 else:
3362 if slot_number is not None:
3363 self._pipeline_slots.add(slot_number)
3365 return self.pipeline_execute_command(*args, **kwargs)
3367 def _validate_watch(self):
3368 if self._explicit_transaction:
3369 raise RedisError("Cannot issue a WATCH after a MULTI")
3371 self._watching = True
3373 def _immediate_execute_command(self, *args, **options):
3374 return self._retry.call_with_retry(
3375 lambda: self._get_connection_and_send_command(*args, **options),
3376 self._reinitialize_on_error,
3377 )
3379 def _get_connection_and_send_command(self, *args, **options):
3380 redis_node, connection = self._get_client_and_connection_for_transaction()
3381 return self._send_command_parse_response(
3382 connection, redis_node, args[0], *args, **options
3383 )
3385 def _send_command_parse_response(
3386 self, conn, redis_node: Redis, command_name, *args, **options
3387 ):
3388 """
3389 Send a command and parse the response
3390 """
3392 conn.send_command(*args)
3393 output = redis_node.parse_response(conn, command_name, **options)
3395 if command_name in self.UNWATCH_COMMANDS:
3396 self._watching = False
3397 return output
3399 def _reinitialize_on_error(self, error):
3400 if self._watching:
3401 if type(error) in self.SLOT_REDIRECT_ERRORS and self._executing:
3402 raise WatchError("Slot rebalancing occurred while watching keys")
3404 if (
3405 type(error) in self.SLOT_REDIRECT_ERRORS
3406 or type(error) in self.CONNECTION_ERRORS
3407 ):
3408 if self._transaction_connection:
3409 self._transaction_connection = None
3411 self._pipe.reinitialize_counter += 1
3412 if self._pipe._should_reinitialized():
3413 self._nodes_manager.initialize()
3414 self.reinitialize_counter = 0
3415 else:
3416 if isinstance(error, AskError):
3417 self._nodes_manager.update_moved_exception(error)
3419 self._executing = False
3421 def _raise_first_error(self, responses, stack):
3422 """
3423 Raise the first exception on the stack
3424 """
3425 for r, cmd in zip(responses, stack):
3426 if isinstance(r, Exception):
3427 self.annotate_exception(r, cmd.position + 1, cmd.args)
3428 raise r
3430 def execute(self, raise_on_error: bool = True) -> List[Any]:
3431 stack = self._command_queue
3432 if not stack and (not self._watching or not self._pipeline_slots):
3433 return []
3435 return self._execute_transaction_with_retries(stack, raise_on_error)
3437 def _execute_transaction_with_retries(
3438 self, stack: List["PipelineCommand"], raise_on_error: bool
3439 ):
3440 return self._retry.call_with_retry(
3441 lambda: self._execute_transaction(stack, raise_on_error),
3442 self._reinitialize_on_error,
3443 )
3445 def _execute_transaction(
3446 self, stack: List["PipelineCommand"], raise_on_error: bool
3447 ):
3448 if len(self._pipeline_slots) > 1:
3449 raise CrossSlotTransactionError(
3450 "All keys involved in a cluster transaction must map to the same slot"
3451 )
3453 self._executing = True
3455 redis_node, connection = self._get_client_and_connection_for_transaction()
3457 stack = chain(
3458 [PipelineCommand(("MULTI",))],
3459 stack,
3460 [PipelineCommand(("EXEC",))],
3461 )
3462 commands = [c.args for c in stack if EMPTY_RESPONSE not in c.options]
3463 packed_commands = connection.pack_commands(commands)
3464 connection.send_packed_command(packed_commands)
3465 errors = []
3467 # parse off the response for MULTI
3468 # NOTE: we need to handle ResponseErrors here and continue
3469 # so that we read all the additional command messages from
3470 # the socket
3471 try:
3472 redis_node.parse_response(connection, "MULTI")
3473 except ResponseError as e:
3474 self.annotate_exception(e, 0, "MULTI")
3475 errors.append(e)
3476 except self.CONNECTION_ERRORS as cluster_error:
3477 self.annotate_exception(cluster_error, 0, "MULTI")
3478 raise
3480 # and all the other commands
3481 for i, command in enumerate(self._command_queue):
3482 if EMPTY_RESPONSE in command.options:
3483 errors.append((i, command.options[EMPTY_RESPONSE]))
3484 else:
3485 try:
3486 _ = redis_node.parse_response(connection, "_")
3487 except self.SLOT_REDIRECT_ERRORS as slot_error:
3488 self.annotate_exception(slot_error, i + 1, command.args)
3489 errors.append(slot_error)
3490 except self.CONNECTION_ERRORS as cluster_error:
3491 self.annotate_exception(cluster_error, i + 1, command.args)
3492 raise
3493 except ResponseError as e:
3494 self.annotate_exception(e, i + 1, command.args)
3495 errors.append(e)
3497 response = None
3498 # parse the EXEC.
3499 try:
3500 response = redis_node.parse_response(connection, "EXEC")
3501 except ExecAbortError:
3502 if errors:
3503 raise errors[0]
3504 raise
3506 self._executing = False
3508 # EXEC clears any watched keys
3509 self._watching = False
3511 if response is None:
3512 raise WatchError("Watched variable changed.")
3514 # put any parse errors into the response
3515 for i, e in errors:
3516 response.insert(i, e)
3518 if len(response) != len(self._command_queue):
3519 raise InvalidPipelineStack(
3520 "Unexpected response length for cluster pipeline EXEC."
3521 " Command stack was {} but response had length {}".format(
3522 [c.args[0] for c in self._command_queue], len(response)
3523 )
3524 )
3526 # find any errors in the response and raise if necessary
3527 if raise_on_error or len(errors) > 0:
3528 self._raise_first_error(
3529 response,
3530 self._command_queue,
3531 )
3533 # We have to run response callbacks manually
3534 data = []
3535 for r, cmd in zip(response, self._command_queue):
3536 if not isinstance(r, Exception):
3537 command_name = cmd.args[0]
3538 if command_name in self._pipe.cluster_response_callbacks:
3539 r = self._pipe.cluster_response_callbacks[command_name](
3540 r, **cmd.options
3541 )
3542 data.append(r)
3543 return data
3545 def reset(self):
3546 self._command_queue = []
3548 # make sure to reset the connection state in the event that we were
3549 # watching something
3550 if self._transaction_connection:
3551 try:
3552 if self._watching:
3553 # call this manually since our unwatch or
3554 # immediate_execute_command methods can call reset()
3555 self._transaction_connection.send_command("UNWATCH")
3556 self._transaction_connection.read_response()
3557 # we can safely return the connection to the pool here since we're
3558 # sure we're no longer WATCHing anything
3559 node = self._nodes_manager.find_connection_owner(
3560 self._transaction_connection
3561 )
3562 node.redis_connection.connection_pool.release(
3563 self._transaction_connection
3564 )
3565 self._transaction_connection = None
3566 except self.CONNECTION_ERRORS:
3567 # disconnect will also remove any previous WATCHes
3568 if self._transaction_connection:
3569 self._transaction_connection.disconnect()
3571 # clean up the other instance attributes
3572 self._watching = False
3573 self._explicit_transaction = False
3574 self._pipeline_slots = set()
3575 self._executing = False
3577 def send_cluster_commands(
3578 self, stack, raise_on_error=True, allow_redirections=True
3579 ):
3580 raise NotImplementedError(
3581 "send_cluster_commands cannot be executed in transactional context."
3582 )
3584 def multi(self):
3585 if self._explicit_transaction:
3586 raise RedisError("Cannot issue nested calls to MULTI")
3587 if self._command_queue:
3588 raise RedisError(
3589 "Commands without an initial WATCH have already been issued"
3590 )
3591 self._explicit_transaction = True
3593 def watch(self, *names):
3594 if self._explicit_transaction:
3595 raise RedisError("Cannot issue a WATCH after a MULTI")
3597 return self.execute_command("WATCH", *names)
3599 def unwatch(self):
3600 if self._watching:
3601 return self.execute_command("UNWATCH")
3603 return True
3605 def discard(self):
3606 self.reset()
3608 def delete(self, *names):
3609 return self.execute_command("DEL", *names)
3611 def unlink(self, *names):
3612 return self.execute_command("UNLINK", *names)