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

1421 statements  

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 

12 

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) 

66 

67 

68def get_node_name(host: str, port: Union[str, int]) -> str: 

69 return f"{host}:{port}" 

70 

71 

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() 

79 

80 

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 

88 

89 return cursors, ret 

90 

91 

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 

100 

101 ret_numsub = [(channel, numsub) for channel, numsub in numsub_d.items()] 

102 return ret_numsub 

103 

104 

105def parse_cluster_slots( 

106 resp: Any, **options: Any 

107) -> Dict[Tuple[int, int], Dict[str, Any]]: 

108 current_host = options.get("current_host", "") 

109 

110 def fix_server(*args: Any) -> Tuple[str, Any]: 

111 return str_if_bytes(args[0]) or current_host, args[1] 

112 

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 } 

121 

122 return slots 

123 

124 

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) 

143 

144 return shards 

145 

146 

147def parse_cluster_myshardid(resp, **options): 

148 """ 

149 Parse CLUSTER MYSHARDID response. 

150 """ 

151 return resp.decode("utf-8") 

152 

153 

154PRIMARY = "primary" 

155REPLICA = "replica" 

156SLOT_ID = "slot-id" 

157 

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") 

202 

203 

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 } 

213 

214 return connection_kwargs 

215 

216 

217class AbstractRedisCluster: 

218 RedisClusterRequestTTL = 16 

219 

220 PRIMARIES = "primaries" 

221 REPLICAS = "replicas" 

222 ALL_NODES = "all" 

223 RANDOM = "random" 

224 DEFAULT_NODE = "default-node" 

225 

226 NODE_FLAGS = {PRIMARIES, REPLICAS, ALL_NODES, RANDOM, DEFAULT_NODE} 

227 

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 ) 

348 

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 ) 

384 

385 CLUSTER_COMMANDS_RESPONSE_CALLBACKS = { 

386 "CLUSTER SLOTS": parse_cluster_slots, 

387 "CLUSTER SHARDS": parse_cluster_shards, 

388 "CLUSTER MYSHARDID": parse_cluster_myshardid, 

389 } 

390 

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 ) 

431 

432 ERRORS_ALLOW_RETRY = ( 

433 ConnectionError, 

434 TimeoutError, 

435 ClusterDownError, 

436 SlotNotCoveredError, 

437 ) 

438 

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. 

444 

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) 

462 

463 

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 

469 

470 For example:: 

471 

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] 

475 

476 Three URL schemes are supported: 

477 

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. 

483 

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. 

487 

488 There are several ways to specify a database number. The first value 

489 found will be used: 

490 

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. 

495 

496 If none of these options are specified, the default db=0 is used. 

497 

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. 

505 

506 """ 

507 return cls(url=url, **kwargs) 

508 

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. 

543 

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! 

581 

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. 

608 

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 = [] 

620 

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 ) 

626 

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 ) 

635 

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 ) 

683 

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") 

692 

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 ) 

715 

716 self.cluster_response_callbacks = CaseInsensitiveDict( 

717 self.__class__.CLUSTER_COMMANDS_RESPONSE_CALLBACKS 

718 ) 

719 self.result_callbacks = CaseInsensitiveDict(self.__class__.RESULT_CALLBACKS) 

720 

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 } 

730 

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 } 

749 

750 self._policy_resolver = policy_resolver 

751 self.commands_parser = CommandsParser(self) 

752 

753 # Node where FT.AGGREGATE command is executed. 

754 self._aggregate_nodes = None 

755 self._lock = threading.RLock() 

756 

757 def __enter__(self): 

758 return self 

759 

760 def __exit__(self, exc_type, exc_value, traceback): 

761 self.close() 

762 

763 def __del__(self): 

764 try: 

765 self.close() 

766 except Exception: 

767 pass 

768 

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 

777 

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() 

784 

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") 

794 

795 if self.user_on_connect_func is not None: 

796 self.user_on_connect_func(connection) 

797 

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 

804 

805 def get_node(self, host=None, port=None, node_name=None): 

806 return self.nodes_manager.get_node(host, port, node_name) 

807 

808 def get_primaries(self): 

809 return self.nodes_manager.get_nodes_by_server_type(PRIMARY) 

810 

811 def get_replicas(self): 

812 return self.nodes_manager.get_nodes_by_server_type(REPLICA) 

813 

814 def get_random_node(self): 

815 return random.choice(list(self.nodes_manager.nodes_cache.values())) 

816 

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() 

823 

824 return self.get_random_primary_node() 

825 

826 def get_nodes(self): 

827 return list(self.nodes_manager.nodes_cache.values()) 

828 

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 

846 

847 return slot_cache[node_idx] 

848 

849 def get_default_node(self): 

850 """ 

851 Get the cluster's default node 

852 """ 

853 return self.nodes_manager.default_node 

854 

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] 

867 

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 = [] 

874 

875 for key in keys: 

876 commands.append( 

877 { 

878 "args": (args[0], key), 

879 "kwargs": kwargs, 

880 } 

881 ) 

882 

883 return commands 

884 

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 ) 

893 

894 return self._aggregate_nodes 

895 

896 def get_random_primary_node(self) -> "ClusterNode": 

897 """ 

898 Returns a random primary node 

899 """ 

900 return random.choice(self.get_primaries()) 

901 

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 

907 

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) 

923 

924 return first_successful_response 

925 

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 

936 

937 def set_retry(self, retry: Retry) -> None: 

938 self.retry = retry 

939 

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() 

956 

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) 

963 

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") 

975 

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 ) 

989 

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. 

1004 

1005 If specified, ``timeout`` indicates a maximum life for the lock. 

1006 By default, it will remain locked until release() is called. 

1007 

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. 

1011 

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``. 

1017 

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. 

1022 

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. 

1027 

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: 

1032 

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. 

1045 

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. 

1050 

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 ) 

1070 

1071 def set_response_callback(self, command, callback): 

1072 """Set a custom Response Callback""" 

1073 self.cluster_response_callbacks[command] = callback 

1074 

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() 

1084 

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) 

1092 

1093 if command_flag in self._command_flags_mapping: 

1094 request_policy = self._command_flags_mapping[command_flag] 

1095 

1096 policy_callback = self._policies_callback_mapping[request_policy] 

1097 

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() 

1106 

1107 if args[0].lower() == "ft.aggregate": 

1108 self._aggregate_nodes = nodes 

1109 

1110 return nodes 

1111 

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 

1121 

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) 

1129 

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. 

1134 

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 

1139 

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) 

1144 

1145 def determine_slot(self, *args) -> Optional[int]: 

1146 """ 

1147 Figure out what slot to use based on args. 

1148 

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] 

1157 

1158 # Get the keys in the command 

1159 

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 

1165 

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 ) 

1193 

1194 # single key command 

1195 if len(keys) == 1: 

1196 return self.keyslot(keys[0]) 

1197 

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 ) 

1205 

1206 return slots.pop() 

1207 

1208 def get_encoder(self): 

1209 """ 

1210 Get the connections' encoder 

1211 """ 

1212 return self.encoder 

1213 

1214 def get_connection_kwargs(self): 

1215 """ 

1216 Get the connections' key-word arguments 

1217 """ 

1218 return self.nodes_manager.connection_kwargs 

1219 

1220 def _is_nodes_flag(self, target_nodes): 

1221 return isinstance(target_nodes, str) and target_nodes in self.node_flags 

1222 

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 

1242 

1243 def execute_command(self, *args, **kwargs): 

1244 return self._internal_execute_command(*args, **kwargs) 

1245 

1246 def _internal_execute_command(self, *args, **kwargs): 

1247 """ 

1248 Wrapper for ERRORS_ALLOW_RETRY error handling. 

1249 

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. 

1253 

1254 If it reaches the number of times, the command will raise the exception 

1255 

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()) 

1267 

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 

1271 

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() 

1276 

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() 

1302 

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) 

1335 

1336 if command_policies.response_policy == ResponsePolicy.ONE_SUCCEEDED: 

1337 break 

1338 

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 

1358 

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) 

1370 

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 

1388 

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) 

1397 

1398 # Remove keys entry, it needs only for cache. 

1399 kwargs.pop("keys", None) 

1400 

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() 

1420 

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 

1455 

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 

1460 

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) 

1473 

1474 raise ClusterError("TTL exhausted.") 

1475 

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 

1484 

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. 

1489 

1490 :type command: str 

1491 :type res: dict 

1492 

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] 

1502 

1503 return self._policies_callback_mapping[response_policy](res) 

1504 

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. 

1509 

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) 

1514 

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 

1536 

1537 

1538class ClusterNode: 

1539 def __init__(self, host, port, server_type=None, redis_connection=None): 

1540 if host == "localhost": 

1541 host = socket.gethostbyname(host) 

1542 

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 

1548 

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 ) 

1557 

1558 def __eq__(self, obj): 

1559 return isinstance(obj, ClusterNode) and obj.name == self.name 

1560 

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 

1568 

1569 

1570class LoadBalancingStrategy(Enum): 

1571 ROUND_ROBIN = "round_robin" 

1572 ROUND_ROBIN_REPLICAS = "round_robin_replicas" 

1573 RANDOM_REPLICA = "random_replica" 

1574 

1575 

1576class LoadBalancer: 

1577 """ 

1578 Round-Robin Load Balancing 

1579 """ 

1580 

1581 def __init__(self, start_index: int = 0) -> None: 

1582 self.primary_to_idx = {} 

1583 self.start_index = start_index 

1584 

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 ) 

1599 

1600 def reset(self) -> None: 

1601 self.primary_to_idx.clear() 

1602 

1603 def _get_random_replica_index(self, list_size: int) -> int: 

1604 return random.randint(1, list_size - 1) 

1605 

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 

1616 

1617 

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() 

1661 

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 

1677 

1678 def update_moved_exception(self, exception): 

1679 self._moved_exception = exception 

1680 

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 if redirected_node in self.slots_cache[e.slot_id]: 

1697 # The MOVED error resulted from a failover, and the new slot owner 

1698 # had previously been a replica. 

1699 old_primary = self.slots_cache[e.slot_id][0] 

1700 # Update the old primary to be a replica and add it to the end of 

1701 # the slot's node list 

1702 old_primary.server_type = REPLICA 

1703 self.slots_cache[e.slot_id].append(old_primary) 

1704 # Remove the old replica, which is now a primary, from the slot's 

1705 # node list 

1706 self.slots_cache[e.slot_id].remove(redirected_node) 

1707 # Override the old primary with the new one 

1708 self.slots_cache[e.slot_id][0] = redirected_node 

1709 if self.default_node == old_primary: 

1710 # Update the default node with the new primary 

1711 self.default_node = redirected_node 

1712 else: 

1713 # The new slot owner is a new server, or a server from a different 

1714 # shard. We need to remove all current nodes from the slot's list 

1715 # (including replications) and add just the new node. 

1716 self.slots_cache[e.slot_id] = [redirected_node] 

1717 # Reset moved_exception 

1718 self._moved_exception = None 

1719 

1720 @deprecated_args( 

1721 args_to_warn=["server_type"], 

1722 reason=( 

1723 "In case you need select some load balancing strategy " 

1724 "that will use replicas, please set it through 'load_balancing_strategy'" 

1725 ), 

1726 version="5.3.0", 

1727 ) 

1728 def get_node_from_slot( 

1729 self, 

1730 slot, 

1731 read_from_replicas=False, 

1732 load_balancing_strategy=None, 

1733 server_type=None, 

1734 ) -> ClusterNode: 

1735 """ 

1736 Gets a node that servers this hash slot 

1737 """ 

1738 if self._moved_exception: 

1739 with self._lock: 

1740 if self._moved_exception: 

1741 self._update_moved_slots() 

1742 

1743 if self.slots_cache.get(slot) is None or len(self.slots_cache[slot]) == 0: 

1744 raise SlotNotCoveredError( 

1745 f'Slot "{slot}" not covered by the cluster. ' 

1746 f'"require_full_coverage={self._require_full_coverage}"' 

1747 ) 

1748 

1749 if read_from_replicas is True and load_balancing_strategy is None: 

1750 load_balancing_strategy = LoadBalancingStrategy.ROUND_ROBIN 

1751 

1752 if len(self.slots_cache[slot]) > 1 and load_balancing_strategy: 

1753 # get the server index using the strategy defined in load_balancing_strategy 

1754 primary_name = self.slots_cache[slot][0].name 

1755 node_idx = self.read_load_balancer.get_server_index( 

1756 primary_name, len(self.slots_cache[slot]), load_balancing_strategy 

1757 ) 

1758 elif ( 

1759 server_type is None 

1760 or server_type == PRIMARY 

1761 or len(self.slots_cache[slot]) == 1 

1762 ): 

1763 # return a primary 

1764 node_idx = 0 

1765 else: 

1766 # return a replica 

1767 # randomly choose one of the replicas 

1768 node_idx = random.randint(1, len(self.slots_cache[slot]) - 1) 

1769 

1770 return self.slots_cache[slot][node_idx] 

1771 

1772 def get_nodes_by_server_type(self, server_type): 

1773 """ 

1774 Get all nodes with the specified server type 

1775 :param server_type: 'primary' or 'replica' 

1776 :return: list of ClusterNode 

1777 """ 

1778 return [ 

1779 node 

1780 for node in self.nodes_cache.values() 

1781 if node.server_type == server_type 

1782 ] 

1783 

1784 def populate_startup_nodes(self, nodes): 

1785 """ 

1786 Populate all startup nodes and filters out any duplicates 

1787 """ 

1788 for n in nodes: 

1789 self.startup_nodes[n.name] = n 

1790 

1791 def check_slots_coverage(self, slots_cache): 

1792 # Validate if all slots are covered or if we should try next 

1793 # startup node 

1794 for i in range(0, REDIS_CLUSTER_HASH_SLOTS): 

1795 if i not in slots_cache: 

1796 return False 

1797 return True 

1798 

1799 def create_redis_connections(self, nodes): 

1800 """ 

1801 This function will create a redis connection to all nodes in :nodes: 

1802 """ 

1803 connection_pools = [] 

1804 for node in nodes: 

1805 if node.redis_connection is None: 

1806 node.redis_connection = self.create_redis_node( 

1807 host=node.host, port=node.port, **self.connection_kwargs 

1808 ) 

1809 connection_pools.append(node.redis_connection.connection_pool) 

1810 

1811 self._event_dispatcher.dispatch( 

1812 AfterPooledConnectionsInstantiationEvent( 

1813 connection_pools, ClientType.SYNC, self._credential_provider 

1814 ) 

1815 ) 

1816 

1817 def create_redis_node(self, host, port, **kwargs): 

1818 # We are configuring the connection pool not to retry 

1819 # connections on lower level clients to avoid retrying 

1820 # connections to nodes that are not reachable 

1821 # and to avoid blocking the connection pool. 

1822 # The only error that will have some handling in the lower 

1823 # level clients is ConnectionError which will trigger disconnection 

1824 # of the socket. 

1825 # The retries will be handled on cluster client level 

1826 # where we will have proper handling of the cluster topology 

1827 node_retry_config = Retry( 

1828 backoff=NoBackoff(), retries=0, supported_errors=(ConnectionError,) 

1829 ) 

1830 

1831 protocol = kwargs.get("protocol", None) 

1832 if protocol in [3, "3"]: 

1833 kwargs.update( 

1834 {"maint_notifications_config": MaintNotificationsConfig(enabled=False)} 

1835 ) 

1836 if self.from_url: 

1837 # Create a redis node with a costumed connection pool 

1838 kwargs.update({"host": host}) 

1839 kwargs.update({"port": port}) 

1840 kwargs.update({"cache": self._cache}) 

1841 kwargs.update({"retry": node_retry_config}) 

1842 r = Redis(connection_pool=self.connection_pool_class(**kwargs)) 

1843 else: 

1844 r = Redis( 

1845 host=host, 

1846 port=port, 

1847 cache=self._cache, 

1848 retry=node_retry_config, 

1849 **kwargs, 

1850 ) 

1851 return r 

1852 

1853 def _get_or_create_cluster_node(self, host, port, role, tmp_nodes_cache): 

1854 node_name = get_node_name(host, port) 

1855 # check if we already have this node in the tmp_nodes_cache 

1856 target_node = tmp_nodes_cache.get(node_name) 

1857 if target_node is None: 

1858 # before creating a new cluster node, check if the cluster node already 

1859 # exists in the current nodes cache and has a valid connection so we can 

1860 # reuse it 

1861 target_node = self.nodes_cache.get(node_name) 

1862 if target_node is None or target_node.redis_connection is None: 

1863 # create new cluster node for this cluster 

1864 target_node = ClusterNode(host, port, role) 

1865 if target_node.server_type != role: 

1866 target_node.server_type = role 

1867 # add this node to the nodes cache 

1868 tmp_nodes_cache[target_node.name] = target_node 

1869 

1870 return target_node 

1871 

1872 def initialize(self): 

1873 """ 

1874 Initializes the nodes cache, slots cache and redis connections. 

1875 :startup_nodes: 

1876 Responsible for discovering other nodes in the cluster 

1877 """ 

1878 self.reset() 

1879 tmp_nodes_cache = {} 

1880 tmp_slots = {} 

1881 disagreements = [] 

1882 startup_nodes_reachable = False 

1883 fully_covered = False 

1884 kwargs = self.connection_kwargs 

1885 exception = None 

1886 # Convert to tuple to prevent RuntimeError if self.startup_nodes 

1887 # is modified during iteration 

1888 for startup_node in tuple(self.startup_nodes.values()): 

1889 try: 

1890 if startup_node.redis_connection: 

1891 r = startup_node.redis_connection 

1892 else: 

1893 # Create a new Redis connection 

1894 r = self.create_redis_node( 

1895 startup_node.host, startup_node.port, **kwargs 

1896 ) 

1897 self.startup_nodes[startup_node.name].redis_connection = r 

1898 # Make sure cluster mode is enabled on this node 

1899 try: 

1900 cluster_slots = str_if_bytes(r.execute_command("CLUSTER SLOTS")) 

1901 r.connection_pool.disconnect() 

1902 except ResponseError: 

1903 raise RedisClusterException( 

1904 "Cluster mode is not enabled on this node" 

1905 ) 

1906 startup_nodes_reachable = True 

1907 except Exception as e: 

1908 # Try the next startup node. 

1909 # The exception is saved and raised only if we have no more nodes. 

1910 exception = e 

1911 continue 

1912 

1913 # CLUSTER SLOTS command results in the following output: 

1914 # [[slot_section[from_slot,to_slot,master,replica1,...,replicaN]]] 

1915 # where each node contains the following list: [IP, port, node_id] 

1916 # Therefore, cluster_slots[0][2][0] will be the IP address of the 

1917 # primary node of the first slot section. 

1918 # If there's only one server in the cluster, its ``host`` is '' 

1919 # Fix it to the host in startup_nodes 

1920 if ( 

1921 len(cluster_slots) == 1 

1922 and len(cluster_slots[0][2][0]) == 0 

1923 and len(self.startup_nodes) == 1 

1924 ): 

1925 cluster_slots[0][2][0] = startup_node.host 

1926 

1927 for slot in cluster_slots: 

1928 primary_node = slot[2] 

1929 host = str_if_bytes(primary_node[0]) 

1930 if host == "": 

1931 host = startup_node.host 

1932 port = int(primary_node[1]) 

1933 host, port = self.remap_host_port(host, port) 

1934 

1935 nodes_for_slot = [] 

1936 

1937 target_node = self._get_or_create_cluster_node( 

1938 host, port, PRIMARY, tmp_nodes_cache 

1939 ) 

1940 nodes_for_slot.append(target_node) 

1941 

1942 replica_nodes = slot[3:] 

1943 for replica_node in replica_nodes: 

1944 host = str_if_bytes(replica_node[0]) 

1945 port = int(replica_node[1]) 

1946 host, port = self.remap_host_port(host, port) 

1947 target_replica_node = self._get_or_create_cluster_node( 

1948 host, port, REPLICA, tmp_nodes_cache 

1949 ) 

1950 nodes_for_slot.append(target_replica_node) 

1951 

1952 for i in range(int(slot[0]), int(slot[1]) + 1): 

1953 if i not in tmp_slots: 

1954 tmp_slots[i] = nodes_for_slot 

1955 else: 

1956 # Validate that 2 nodes want to use the same slot cache 

1957 # setup 

1958 tmp_slot = tmp_slots[i][0] 

1959 if tmp_slot.name != target_node.name: 

1960 disagreements.append( 

1961 f"{tmp_slot.name} vs {target_node.name} on slot: {i}" 

1962 ) 

1963 

1964 if len(disagreements) > 5: 

1965 raise RedisClusterException( 

1966 f"startup_nodes could not agree on a valid " 

1967 f"slots cache: {', '.join(disagreements)}" 

1968 ) 

1969 

1970 fully_covered = self.check_slots_coverage(tmp_slots) 

1971 if fully_covered: 

1972 # Don't need to continue to the next startup node if all 

1973 # slots are covered 

1974 break 

1975 

1976 if not startup_nodes_reachable: 

1977 raise RedisClusterException( 

1978 f"Redis Cluster cannot be connected. Please provide at least " 

1979 f"one reachable node: {str(exception)}" 

1980 ) from exception 

1981 

1982 if self._cache is None and self._cache_config is not None: 

1983 if self._cache_factory is None: 

1984 self._cache = CacheFactory(self._cache_config).get_cache() 

1985 else: 

1986 self._cache = self._cache_factory.get_cache() 

1987 

1988 # Create Redis connections to all nodes 

1989 self.create_redis_connections(list(tmp_nodes_cache.values())) 

1990 

1991 # Check if the slots are not fully covered 

1992 if not fully_covered and self._require_full_coverage: 

1993 # Despite the requirement that the slots be covered, there 

1994 # isn't a full coverage 

1995 raise RedisClusterException( 

1996 f"All slots are not covered after query all startup_nodes. " 

1997 f"{len(tmp_slots)} of {REDIS_CLUSTER_HASH_SLOTS} " 

1998 f"covered..." 

1999 ) 

2000 

2001 # Set the tmp variables to the real variables 

2002 self.nodes_cache = tmp_nodes_cache 

2003 self.slots_cache = tmp_slots 

2004 # Set the default node 

2005 self.default_node = self.get_nodes_by_server_type(PRIMARY)[0] 

2006 if self._dynamic_startup_nodes: 

2007 # Populate the startup nodes with all discovered nodes 

2008 self.startup_nodes = tmp_nodes_cache 

2009 # If initialize was called after a MovedError, clear it 

2010 self._moved_exception = None 

2011 

2012 def close(self) -> None: 

2013 self.default_node = None 

2014 for node in self.nodes_cache.values(): 

2015 if node.redis_connection: 

2016 node.redis_connection.close() 

2017 

2018 def reset(self): 

2019 try: 

2020 self.read_load_balancer.reset() 

2021 except TypeError: 

2022 # The read_load_balancer is None, do nothing 

2023 pass 

2024 

2025 def remap_host_port(self, host: str, port: int) -> Tuple[str, int]: 

2026 """ 

2027 Remap the host and port returned from the cluster to a different 

2028 internal value. Useful if the client is not connecting directly 

2029 to the cluster. 

2030 """ 

2031 if self.address_remap: 

2032 return self.address_remap((host, port)) 

2033 return host, port 

2034 

2035 def find_connection_owner(self, connection: Connection) -> Optional[Redis]: 

2036 node_name = get_node_name(connection.host, connection.port) 

2037 for node in tuple(self.nodes_cache.values()): 

2038 if node.redis_connection: 

2039 conn_args = node.redis_connection.connection_pool.connection_kwargs 

2040 if node_name == get_node_name( 

2041 conn_args.get("host"), conn_args.get("port") 

2042 ): 

2043 return node 

2044 

2045 

2046class ClusterPubSub(PubSub): 

2047 """ 

2048 Wrapper for PubSub class. 

2049 

2050 IMPORTANT: before using ClusterPubSub, read about the known limitations 

2051 with pubsub in Cluster mode and learn how to workaround them: 

2052 https://redis-py-cluster.readthedocs.io/en/stable/pubsub.html 

2053 """ 

2054 

2055 def __init__( 

2056 self, 

2057 redis_cluster, 

2058 node=None, 

2059 host=None, 

2060 port=None, 

2061 push_handler_func=None, 

2062 event_dispatcher: Optional["EventDispatcher"] = None, 

2063 **kwargs, 

2064 ): 

2065 """ 

2066 When a pubsub instance is created without specifying a node, a single 

2067 node will be transparently chosen for the pubsub connection on the 

2068 first command execution. The node will be determined by: 

2069 1. Hashing the channel name in the request to find its keyslot 

2070 2. Selecting a node that handles the keyslot: If read_from_replicas is 

2071 set to true or load_balancing_strategy is set, a replica can be selected. 

2072 

2073 :type redis_cluster: RedisCluster 

2074 :type node: ClusterNode 

2075 :type host: str 

2076 :type port: int 

2077 """ 

2078 self.node = None 

2079 self.set_pubsub_node(redis_cluster, node, host, port) 

2080 connection_pool = ( 

2081 None 

2082 if self.node is None 

2083 else redis_cluster.get_redis_connection(self.node).connection_pool 

2084 ) 

2085 self.cluster = redis_cluster 

2086 self.node_pubsub_mapping = {} 

2087 self._pubsubs_generator = self._pubsubs_generator() 

2088 if event_dispatcher is None: 

2089 self._event_dispatcher = EventDispatcher() 

2090 else: 

2091 self._event_dispatcher = event_dispatcher 

2092 super().__init__( 

2093 connection_pool=connection_pool, 

2094 encoder=redis_cluster.encoder, 

2095 push_handler_func=push_handler_func, 

2096 event_dispatcher=self._event_dispatcher, 

2097 **kwargs, 

2098 ) 

2099 

2100 def set_pubsub_node(self, cluster, node=None, host=None, port=None): 

2101 """ 

2102 The pubsub node will be set according to the passed node, host and port 

2103 When none of the node, host, or port are specified - the node is set 

2104 to None and will be determined by the keyslot of the channel in the 

2105 first command to be executed. 

2106 RedisClusterException will be thrown if the passed node does not exist 

2107 in the cluster. 

2108 If host is passed without port, or vice versa, a DataError will be 

2109 thrown. 

2110 :type cluster: RedisCluster 

2111 :type node: ClusterNode 

2112 :type host: str 

2113 :type port: int 

2114 """ 

2115 if node is not None: 

2116 # node is passed by the user 

2117 self._raise_on_invalid_node(cluster, node, node.host, node.port) 

2118 pubsub_node = node 

2119 elif host is not None and port is not None: 

2120 # host and port passed by the user 

2121 node = cluster.get_node(host=host, port=port) 

2122 self._raise_on_invalid_node(cluster, node, host, port) 

2123 pubsub_node = node 

2124 elif any([host, port]) is True: 

2125 # only 'host' or 'port' passed 

2126 raise DataError("Passing a host requires passing a port, and vice versa") 

2127 else: 

2128 # nothing passed by the user. set node to None 

2129 pubsub_node = None 

2130 

2131 self.node = pubsub_node 

2132 

2133 def get_pubsub_node(self): 

2134 """ 

2135 Get the node that is being used as the pubsub connection 

2136 """ 

2137 return self.node 

2138 

2139 def _raise_on_invalid_node(self, redis_cluster, node, host, port): 

2140 """ 

2141 Raise a RedisClusterException if the node is None or doesn't exist in 

2142 the cluster. 

2143 """ 

2144 if node is None or redis_cluster.get_node(node_name=node.name) is None: 

2145 raise RedisClusterException( 

2146 f"Node {host}:{port} doesn't exist in the cluster" 

2147 ) 

2148 

2149 def execute_command(self, *args): 

2150 """ 

2151 Execute a subscribe/unsubscribe command. 

2152 

2153 Taken code from redis-py and tweak to make it work within a cluster. 

2154 """ 

2155 # NOTE: don't parse the response in this function -- it could pull a 

2156 # legitimate message off the stack if the connection is already 

2157 # subscribed to one or more channels 

2158 

2159 if self.connection is None: 

2160 if self.connection_pool is None: 

2161 if len(args) > 1: 

2162 # Hash the first channel and get one of the nodes holding 

2163 # this slot 

2164 channel = args[1] 

2165 slot = self.cluster.keyslot(channel) 

2166 node = self.cluster.nodes_manager.get_node_from_slot( 

2167 slot, 

2168 self.cluster.read_from_replicas, 

2169 self.cluster.load_balancing_strategy, 

2170 ) 

2171 else: 

2172 # Get a random node 

2173 node = self.cluster.get_random_node() 

2174 self.node = node 

2175 redis_connection = self.cluster.get_redis_connection(node) 

2176 self.connection_pool = redis_connection.connection_pool 

2177 self.connection = self.connection_pool.get_connection() 

2178 # register a callback that re-subscribes to any channels we 

2179 # were listening to when we were disconnected 

2180 self.connection.register_connect_callback(self.on_connect) 

2181 if self.push_handler_func is not None: 

2182 self.connection._parser.set_pubsub_push_handler(self.push_handler_func) 

2183 self._event_dispatcher.dispatch( 

2184 AfterPubSubConnectionInstantiationEvent( 

2185 self.connection, self.connection_pool, ClientType.SYNC, self._lock 

2186 ) 

2187 ) 

2188 connection = self.connection 

2189 self._execute(connection, connection.send_command, *args) 

2190 

2191 def _get_node_pubsub(self, node): 

2192 try: 

2193 return self.node_pubsub_mapping[node.name] 

2194 except KeyError: 

2195 pubsub = node.redis_connection.pubsub( 

2196 push_handler_func=self.push_handler_func 

2197 ) 

2198 self.node_pubsub_mapping[node.name] = pubsub 

2199 return pubsub 

2200 

2201 def _sharded_message_generator(self): 

2202 for _ in range(len(self.node_pubsub_mapping)): 

2203 pubsub = next(self._pubsubs_generator) 

2204 message = pubsub.get_message() 

2205 if message is not None: 

2206 return message 

2207 return None 

2208 

2209 def _pubsubs_generator(self): 

2210 while True: 

2211 current_nodes = list(self.node_pubsub_mapping.values()) 

2212 yield from current_nodes 

2213 

2214 def get_sharded_message( 

2215 self, ignore_subscribe_messages=False, timeout=0.0, target_node=None 

2216 ): 

2217 if target_node: 

2218 message = self.node_pubsub_mapping[target_node.name].get_message( 

2219 ignore_subscribe_messages=ignore_subscribe_messages, timeout=timeout 

2220 ) 

2221 else: 

2222 message = self._sharded_message_generator() 

2223 if message is None: 

2224 return None 

2225 elif str_if_bytes(message["type"]) == "sunsubscribe": 

2226 if message["channel"] in self.pending_unsubscribe_shard_channels: 

2227 self.pending_unsubscribe_shard_channels.remove(message["channel"]) 

2228 self.shard_channels.pop(message["channel"], None) 

2229 node = self.cluster.get_node_from_key(message["channel"]) 

2230 if self.node_pubsub_mapping[node.name].subscribed is False: 

2231 self.node_pubsub_mapping.pop(node.name) 

2232 if not self.channels and not self.patterns and not self.shard_channels: 

2233 # There are no subscriptions anymore, set subscribed_event flag 

2234 # to false 

2235 self.subscribed_event.clear() 

2236 if self.ignore_subscribe_messages or ignore_subscribe_messages: 

2237 return None 

2238 return message 

2239 

2240 def ssubscribe(self, *args, **kwargs): 

2241 if args: 

2242 args = list_or_args(args[0], args[1:]) 

2243 s_channels = dict.fromkeys(args) 

2244 s_channels.update(kwargs) 

2245 for s_channel, handler in s_channels.items(): 

2246 node = self.cluster.get_node_from_key(s_channel) 

2247 pubsub = self._get_node_pubsub(node) 

2248 if handler: 

2249 pubsub.ssubscribe(**{s_channel: handler}) 

2250 else: 

2251 pubsub.ssubscribe(s_channel) 

2252 self.shard_channels.update(pubsub.shard_channels) 

2253 self.pending_unsubscribe_shard_channels.difference_update( 

2254 self._normalize_keys({s_channel: None}) 

2255 ) 

2256 if pubsub.subscribed and not self.subscribed: 

2257 self.subscribed_event.set() 

2258 self.health_check_response_counter = 0 

2259 

2260 def sunsubscribe(self, *args): 

2261 if args: 

2262 args = list_or_args(args[0], args[1:]) 

2263 else: 

2264 args = self.shard_channels 

2265 

2266 for s_channel in args: 

2267 node = self.cluster.get_node_from_key(s_channel) 

2268 p = self._get_node_pubsub(node) 

2269 p.sunsubscribe(s_channel) 

2270 self.pending_unsubscribe_shard_channels.update( 

2271 p.pending_unsubscribe_shard_channels 

2272 ) 

2273 

2274 def get_redis_connection(self): 

2275 """ 

2276 Get the Redis connection of the pubsub connected node. 

2277 """ 

2278 if self.node is not None: 

2279 return self.node.redis_connection 

2280 

2281 def disconnect(self): 

2282 """ 

2283 Disconnect the pubsub connection. 

2284 """ 

2285 if self.connection: 

2286 self.connection.disconnect() 

2287 for pubsub in self.node_pubsub_mapping.values(): 

2288 pubsub.connection.disconnect() 

2289 

2290 

2291class ClusterPipeline(RedisCluster): 

2292 """ 

2293 Support for Redis pipeline 

2294 in cluster mode 

2295 """ 

2296 

2297 ERRORS_ALLOW_RETRY = ( 

2298 ConnectionError, 

2299 TimeoutError, 

2300 MovedError, 

2301 AskError, 

2302 TryAgainError, 

2303 ) 

2304 

2305 NO_SLOTS_COMMANDS = {"UNWATCH"} 

2306 IMMEDIATE_EXECUTE_COMMANDS = {"WATCH", "UNWATCH"} 

2307 UNWATCH_COMMANDS = {"DISCARD", "EXEC", "UNWATCH"} 

2308 

2309 @deprecated_args( 

2310 args_to_warn=[ 

2311 "cluster_error_retry_attempts", 

2312 ], 

2313 reason="Please configure the 'retry' object instead", 

2314 version="6.0.0", 

2315 ) 

2316 def __init__( 

2317 self, 

2318 nodes_manager: "NodesManager", 

2319 commands_parser: "CommandsParser", 

2320 result_callbacks: Optional[Dict[str, Callable]] = None, 

2321 cluster_response_callbacks: Optional[Dict[str, Callable]] = None, 

2322 startup_nodes: Optional[List["ClusterNode"]] = None, 

2323 read_from_replicas: bool = False, 

2324 load_balancing_strategy: Optional[LoadBalancingStrategy] = None, 

2325 cluster_error_retry_attempts: int = 3, 

2326 reinitialize_steps: int = 5, 

2327 retry: Optional[Retry] = None, 

2328 lock=None, 

2329 transaction=False, 

2330 policy_resolver: PolicyResolver = StaticPolicyResolver(), 

2331 **kwargs, 

2332 ): 

2333 """ """ 

2334 self.command_stack = [] 

2335 self.nodes_manager = nodes_manager 

2336 self.commands_parser = commands_parser 

2337 self.refresh_table_asap = False 

2338 self.result_callbacks = ( 

2339 result_callbacks or self.__class__.RESULT_CALLBACKS.copy() 

2340 ) 

2341 self.startup_nodes = startup_nodes if startup_nodes else [] 

2342 self.read_from_replicas = read_from_replicas 

2343 self.load_balancing_strategy = load_balancing_strategy 

2344 self.command_flags = self.__class__.COMMAND_FLAGS.copy() 

2345 self.cluster_response_callbacks = cluster_response_callbacks 

2346 self.reinitialize_counter = 0 

2347 self.reinitialize_steps = reinitialize_steps 

2348 if retry is not None: 

2349 self.retry = retry 

2350 else: 

2351 self.retry = Retry( 

2352 backoff=ExponentialWithJitterBackoff(base=1, cap=10), 

2353 retries=cluster_error_retry_attempts, 

2354 ) 

2355 

2356 self.encoder = Encoder( 

2357 kwargs.get("encoding", "utf-8"), 

2358 kwargs.get("encoding_errors", "strict"), 

2359 kwargs.get("decode_responses", False), 

2360 ) 

2361 if lock is None: 

2362 lock = threading.RLock() 

2363 self._lock = lock 

2364 self.parent_execute_command = super().execute_command 

2365 self._execution_strategy: ExecutionStrategy = ( 

2366 PipelineStrategy(self) if not transaction else TransactionStrategy(self) 

2367 ) 

2368 

2369 # For backward compatibility, mapping from existing policies to new one 

2370 self._command_flags_mapping: dict[str, Union[RequestPolicy, ResponsePolicy]] = { 

2371 self.__class__.RANDOM: RequestPolicy.DEFAULT_KEYLESS, 

2372 self.__class__.PRIMARIES: RequestPolicy.ALL_SHARDS, 

2373 self.__class__.ALL_NODES: RequestPolicy.ALL_NODES, 

2374 self.__class__.REPLICAS: RequestPolicy.ALL_REPLICAS, 

2375 self.__class__.DEFAULT_NODE: RequestPolicy.DEFAULT_NODE, 

2376 SLOT_ID: RequestPolicy.DEFAULT_KEYED, 

2377 } 

2378 

2379 self._policies_callback_mapping: dict[ 

2380 Union[RequestPolicy, ResponsePolicy], Callable 

2381 ] = { 

2382 RequestPolicy.DEFAULT_KEYLESS: lambda command_name: [ 

2383 self.get_random_primary_or_all_nodes(command_name) 

2384 ], 

2385 RequestPolicy.DEFAULT_KEYED: lambda command, 

2386 *args: self.get_nodes_from_slot(command, *args), 

2387 RequestPolicy.DEFAULT_NODE: lambda: [self.get_default_node()], 

2388 RequestPolicy.ALL_SHARDS: self.get_primaries, 

2389 RequestPolicy.ALL_NODES: self.get_nodes, 

2390 RequestPolicy.ALL_REPLICAS: self.get_replicas, 

2391 RequestPolicy.MULTI_SHARD: lambda *args, 

2392 **kwargs: self._split_multi_shard_command(*args, **kwargs), 

2393 RequestPolicy.SPECIAL: self.get_special_nodes, 

2394 ResponsePolicy.DEFAULT_KEYLESS: lambda res: res, 

2395 ResponsePolicy.DEFAULT_KEYED: lambda res: res, 

2396 } 

2397 

2398 self._policy_resolver = policy_resolver 

2399 

2400 def __repr__(self): 

2401 """ """ 

2402 return f"{type(self).__name__}" 

2403 

2404 def __enter__(self): 

2405 """ """ 

2406 return self 

2407 

2408 def __exit__(self, exc_type, exc_value, traceback): 

2409 """ """ 

2410 self.reset() 

2411 

2412 def __del__(self): 

2413 try: 

2414 self.reset() 

2415 except Exception: 

2416 pass 

2417 

2418 def __len__(self): 

2419 """ """ 

2420 return len(self._execution_strategy.command_queue) 

2421 

2422 def __bool__(self): 

2423 "Pipeline instances should always evaluate to True on Python 3+" 

2424 return True 

2425 

2426 def execute_command(self, *args, **kwargs): 

2427 """ 

2428 Wrapper function for pipeline_execute_command 

2429 """ 

2430 return self._execution_strategy.execute_command(*args, **kwargs) 

2431 

2432 def pipeline_execute_command(self, *args, **options): 

2433 """ 

2434 Stage a command to be executed when execute() is next called 

2435 

2436 Returns the current Pipeline object back so commands can be 

2437 chained together, such as: 

2438 

2439 pipe = pipe.set('foo', 'bar').incr('baz').decr('bang') 

2440 

2441 At some other point, you can then run: pipe.execute(), 

2442 which will execute all commands queued in the pipe. 

2443 """ 

2444 return self._execution_strategy.execute_command(*args, **options) 

2445 

2446 def annotate_exception(self, exception, number, command): 

2447 """ 

2448 Provides extra context to the exception prior to it being handled 

2449 """ 

2450 self._execution_strategy.annotate_exception(exception, number, command) 

2451 

2452 def execute(self, raise_on_error: bool = True) -> List[Any]: 

2453 """ 

2454 Execute all the commands in the current pipeline 

2455 """ 

2456 

2457 try: 

2458 return self._execution_strategy.execute(raise_on_error) 

2459 finally: 

2460 self.reset() 

2461 

2462 def reset(self): 

2463 """ 

2464 Reset back to empty pipeline. 

2465 """ 

2466 self._execution_strategy.reset() 

2467 

2468 def send_cluster_commands( 

2469 self, stack, raise_on_error=True, allow_redirections=True 

2470 ): 

2471 return self._execution_strategy.send_cluster_commands( 

2472 stack, raise_on_error=raise_on_error, allow_redirections=allow_redirections 

2473 ) 

2474 

2475 def exists(self, *keys): 

2476 return self._execution_strategy.exists(*keys) 

2477 

2478 def eval(self): 

2479 """ """ 

2480 return self._execution_strategy.eval() 

2481 

2482 def multi(self): 

2483 """ 

2484 Start a transactional block of the pipeline after WATCH commands 

2485 are issued. End the transactional block with `execute`. 

2486 """ 

2487 self._execution_strategy.multi() 

2488 

2489 def load_scripts(self): 

2490 """ """ 

2491 self._execution_strategy.load_scripts() 

2492 

2493 def discard(self): 

2494 """ """ 

2495 self._execution_strategy.discard() 

2496 

2497 def watch(self, *names): 

2498 """Watches the values at keys ``names``""" 

2499 self._execution_strategy.watch(*names) 

2500 

2501 def unwatch(self): 

2502 """Unwatches all previously specified keys""" 

2503 self._execution_strategy.unwatch() 

2504 

2505 def script_load_for_pipeline(self, *args, **kwargs): 

2506 self._execution_strategy.script_load_for_pipeline(*args, **kwargs) 

2507 

2508 def delete(self, *names): 

2509 self._execution_strategy.delete(*names) 

2510 

2511 def unlink(self, *names): 

2512 self._execution_strategy.unlink(*names) 

2513 

2514 

2515def block_pipeline_command(name: str) -> Callable[..., Any]: 

2516 """ 

2517 Prints error because some pipelined commands should 

2518 be blocked when running in cluster-mode 

2519 """ 

2520 

2521 def inner(*args, **kwargs): 

2522 raise RedisClusterException( 

2523 f"ERROR: Calling pipelined function {name} is blocked " 

2524 f"when running redis in cluster mode..." 

2525 ) 

2526 

2527 return inner 

2528 

2529 

2530# Blocked pipeline commands 

2531PIPELINE_BLOCKED_COMMANDS = ( 

2532 "BGREWRITEAOF", 

2533 "BGSAVE", 

2534 "BITOP", 

2535 "BRPOPLPUSH", 

2536 "CLIENT GETNAME", 

2537 "CLIENT KILL", 

2538 "CLIENT LIST", 

2539 "CLIENT SETNAME", 

2540 "CLIENT", 

2541 "CONFIG GET", 

2542 "CONFIG RESETSTAT", 

2543 "CONFIG REWRITE", 

2544 "CONFIG SET", 

2545 "CONFIG", 

2546 "DBSIZE", 

2547 "ECHO", 

2548 "EVALSHA", 

2549 "FLUSHALL", 

2550 "FLUSHDB", 

2551 "INFO", 

2552 "KEYS", 

2553 "LASTSAVE", 

2554 "MGET", 

2555 "MGET NONATOMIC", 

2556 "MOVE", 

2557 "MSET", 

2558 "MSETEX", 

2559 "MSET NONATOMIC", 

2560 "MSETNX", 

2561 "PFCOUNT", 

2562 "PFMERGE", 

2563 "PING", 

2564 "PUBLISH", 

2565 "RANDOMKEY", 

2566 "READONLY", 

2567 "READWRITE", 

2568 "RENAME", 

2569 "RENAMENX", 

2570 "RPOPLPUSH", 

2571 "SAVE", 

2572 "SCAN", 

2573 "SCRIPT EXISTS", 

2574 "SCRIPT FLUSH", 

2575 "SCRIPT KILL", 

2576 "SCRIPT LOAD", 

2577 "SCRIPT", 

2578 "SDIFF", 

2579 "SDIFFSTORE", 

2580 "SENTINEL GET MASTER ADDR BY NAME", 

2581 "SENTINEL MASTER", 

2582 "SENTINEL MASTERS", 

2583 "SENTINEL MONITOR", 

2584 "SENTINEL REMOVE", 

2585 "SENTINEL SENTINELS", 

2586 "SENTINEL SET", 

2587 "SENTINEL SLAVES", 

2588 "SENTINEL", 

2589 "SHUTDOWN", 

2590 "SINTER", 

2591 "SINTERSTORE", 

2592 "SLAVEOF", 

2593 "SLOWLOG GET", 

2594 "SLOWLOG LEN", 

2595 "SLOWLOG RESET", 

2596 "SLOWLOG", 

2597 "SMOVE", 

2598 "SORT", 

2599 "SUNION", 

2600 "SUNIONSTORE", 

2601 "TIME", 

2602) 

2603for command in PIPELINE_BLOCKED_COMMANDS: 

2604 command = command.replace(" ", "_").lower() 

2605 

2606 setattr(ClusterPipeline, command, block_pipeline_command(command)) 

2607 

2608 

2609class PipelineCommand: 

2610 """ """ 

2611 

2612 def __init__(self, args, options=None, position=None): 

2613 self.args = args 

2614 if options is None: 

2615 options = {} 

2616 self.options = options 

2617 self.position = position 

2618 self.result = None 

2619 self.node = None 

2620 self.asking = False 

2621 self.command_policies: Optional[CommandPolicies] = None 

2622 

2623 

2624class NodeCommands: 

2625 """ """ 

2626 

2627 def __init__(self, parse_response, connection_pool, connection): 

2628 """ """ 

2629 self.parse_response = parse_response 

2630 self.connection_pool = connection_pool 

2631 self.connection = connection 

2632 self.commands = [] 

2633 

2634 def append(self, c): 

2635 """ """ 

2636 self.commands.append(c) 

2637 

2638 def write(self): 

2639 """ 

2640 Code borrowed from Redis so it can be fixed 

2641 """ 

2642 connection = self.connection 

2643 commands = self.commands 

2644 

2645 # We are going to clobber the commands with the write, so go ahead 

2646 # and ensure that nothing is sitting there from a previous run. 

2647 for c in commands: 

2648 c.result = None 

2649 

2650 # build up all commands into a single request to increase network perf 

2651 # send all the commands and catch connection and timeout errors. 

2652 try: 

2653 connection.send_packed_command( 

2654 connection.pack_commands([c.args for c in commands]) 

2655 ) 

2656 except (ConnectionError, TimeoutError) as e: 

2657 for c in commands: 

2658 c.result = e 

2659 

2660 def read(self): 

2661 """ """ 

2662 connection = self.connection 

2663 for c in self.commands: 

2664 # if there is a result on this command, 

2665 # it means we ran into an exception 

2666 # like a connection error. Trying to parse 

2667 # a response on a connection that 

2668 # is no longer open will result in a 

2669 # connection error raised by redis-py. 

2670 # but redis-py doesn't check in parse_response 

2671 # that the sock object is 

2672 # still set and if you try to 

2673 # read from a closed connection, it will 

2674 # result in an AttributeError because 

2675 # it will do a readline() call on None. 

2676 # This can have all kinds of nasty side-effects. 

2677 # Treating this case as a connection error 

2678 # is fine because it will dump 

2679 # the connection object back into the 

2680 # pool and on the next write, it will 

2681 # explicitly open the connection and all will be well. 

2682 if c.result is None: 

2683 try: 

2684 c.result = self.parse_response(connection, c.args[0], **c.options) 

2685 except (ConnectionError, TimeoutError) as e: 

2686 for c in self.commands: 

2687 c.result = e 

2688 return 

2689 except RedisError: 

2690 c.result = sys.exc_info()[1] 

2691 

2692 

2693class ExecutionStrategy(ABC): 

2694 @property 

2695 @abstractmethod 

2696 def command_queue(self): 

2697 pass 

2698 

2699 @abstractmethod 

2700 def execute_command(self, *args, **kwargs): 

2701 """ 

2702 Execution flow for current execution strategy. 

2703 

2704 See: ClusterPipeline.execute_command() 

2705 """ 

2706 pass 

2707 

2708 @abstractmethod 

2709 def annotate_exception(self, exception, number, command): 

2710 """ 

2711 Annotate exception according to current execution strategy. 

2712 

2713 See: ClusterPipeline.annotate_exception() 

2714 """ 

2715 pass 

2716 

2717 @abstractmethod 

2718 def pipeline_execute_command(self, *args, **options): 

2719 """ 

2720 Pipeline execution flow for current execution strategy. 

2721 

2722 See: ClusterPipeline.pipeline_execute_command() 

2723 """ 

2724 pass 

2725 

2726 @abstractmethod 

2727 def execute(self, raise_on_error: bool = True) -> List[Any]: 

2728 """ 

2729 Executes current execution strategy. 

2730 

2731 See: ClusterPipeline.execute() 

2732 """ 

2733 pass 

2734 

2735 @abstractmethod 

2736 def send_cluster_commands( 

2737 self, stack, raise_on_error=True, allow_redirections=True 

2738 ): 

2739 """ 

2740 Sends commands according to current execution strategy. 

2741 

2742 See: ClusterPipeline.send_cluster_commands() 

2743 """ 

2744 pass 

2745 

2746 @abstractmethod 

2747 def reset(self): 

2748 """ 

2749 Resets current execution strategy. 

2750 

2751 See: ClusterPipeline.reset() 

2752 """ 

2753 pass 

2754 

2755 @abstractmethod 

2756 def exists(self, *keys): 

2757 pass 

2758 

2759 @abstractmethod 

2760 def eval(self): 

2761 pass 

2762 

2763 @abstractmethod 

2764 def multi(self): 

2765 """ 

2766 Starts transactional context. 

2767 

2768 See: ClusterPipeline.multi() 

2769 """ 

2770 pass 

2771 

2772 @abstractmethod 

2773 def load_scripts(self): 

2774 pass 

2775 

2776 @abstractmethod 

2777 def watch(self, *names): 

2778 pass 

2779 

2780 @abstractmethod 

2781 def unwatch(self): 

2782 """ 

2783 Unwatches all previously specified keys 

2784 

2785 See: ClusterPipeline.unwatch() 

2786 """ 

2787 pass 

2788 

2789 @abstractmethod 

2790 def script_load_for_pipeline(self, *args, **kwargs): 

2791 pass 

2792 

2793 @abstractmethod 

2794 def delete(self, *names): 

2795 """ 

2796 "Delete a key specified by ``names``" 

2797 

2798 See: ClusterPipeline.delete() 

2799 """ 

2800 pass 

2801 

2802 @abstractmethod 

2803 def unlink(self, *names): 

2804 """ 

2805 "Unlink a key specified by ``names``" 

2806 

2807 See: ClusterPipeline.unlink() 

2808 """ 

2809 pass 

2810 

2811 @abstractmethod 

2812 def discard(self): 

2813 pass 

2814 

2815 

2816class AbstractStrategy(ExecutionStrategy): 

2817 def __init__( 

2818 self, 

2819 pipe: ClusterPipeline, 

2820 ): 

2821 self._command_queue: List[PipelineCommand] = [] 

2822 self._pipe = pipe 

2823 self._nodes_manager = self._pipe.nodes_manager 

2824 

2825 @property 

2826 def command_queue(self): 

2827 return self._command_queue 

2828 

2829 @command_queue.setter 

2830 def command_queue(self, queue: List[PipelineCommand]): 

2831 self._command_queue = queue 

2832 

2833 @abstractmethod 

2834 def execute_command(self, *args, **kwargs): 

2835 pass 

2836 

2837 def pipeline_execute_command(self, *args, **options): 

2838 self._command_queue.append( 

2839 PipelineCommand(args, options, len(self._command_queue)) 

2840 ) 

2841 return self._pipe 

2842 

2843 @abstractmethod 

2844 def execute(self, raise_on_error: bool = True) -> List[Any]: 

2845 pass 

2846 

2847 @abstractmethod 

2848 def send_cluster_commands( 

2849 self, stack, raise_on_error=True, allow_redirections=True 

2850 ): 

2851 pass 

2852 

2853 @abstractmethod 

2854 def reset(self): 

2855 pass 

2856 

2857 def exists(self, *keys): 

2858 return self.execute_command("EXISTS", *keys) 

2859 

2860 def eval(self): 

2861 """ """ 

2862 raise RedisClusterException("method eval() is not implemented") 

2863 

2864 def load_scripts(self): 

2865 """ """ 

2866 raise RedisClusterException("method load_scripts() is not implemented") 

2867 

2868 def script_load_for_pipeline(self, *args, **kwargs): 

2869 """ """ 

2870 raise RedisClusterException( 

2871 "method script_load_for_pipeline() is not implemented" 

2872 ) 

2873 

2874 def annotate_exception(self, exception, number, command): 

2875 """ 

2876 Provides extra context to the exception prior to it being handled 

2877 """ 

2878 cmd = " ".join(map(safe_str, command)) 

2879 msg = ( 

2880 f"Command # {number} ({truncate_text(cmd)}) of pipeline " 

2881 f"caused error: {exception.args[0]}" 

2882 ) 

2883 exception.args = (msg,) + exception.args[1:] 

2884 

2885 

2886class PipelineStrategy(AbstractStrategy): 

2887 def __init__(self, pipe: ClusterPipeline): 

2888 super().__init__(pipe) 

2889 self.command_flags = pipe.command_flags 

2890 

2891 def execute_command(self, *args, **kwargs): 

2892 return self.pipeline_execute_command(*args, **kwargs) 

2893 

2894 def _raise_first_error(self, stack): 

2895 """ 

2896 Raise the first exception on the stack 

2897 """ 

2898 for c in stack: 

2899 r = c.result 

2900 if isinstance(r, Exception): 

2901 self.annotate_exception(r, c.position + 1, c.args) 

2902 raise r 

2903 

2904 def execute(self, raise_on_error: bool = True) -> List[Any]: 

2905 stack = self._command_queue 

2906 if not stack: 

2907 return [] 

2908 

2909 try: 

2910 return self.send_cluster_commands(stack, raise_on_error) 

2911 finally: 

2912 self.reset() 

2913 

2914 def reset(self): 

2915 """ 

2916 Reset back to empty pipeline. 

2917 """ 

2918 self._command_queue = [] 

2919 

2920 def send_cluster_commands( 

2921 self, stack, raise_on_error=True, allow_redirections=True 

2922 ): 

2923 """ 

2924 Wrapper for RedisCluster.ERRORS_ALLOW_RETRY errors handling. 

2925 

2926 If one of the retryable exceptions has been thrown we assume that: 

2927 - connection_pool was disconnected 

2928 - connection_pool was reset 

2929 - refresh_table_asap set to True 

2930 

2931 It will try the number of times specified by 

2932 the retries in config option "self.retry" 

2933 which defaults to 3 unless manually configured. 

2934 

2935 If it reaches the number of times, the command will 

2936 raises ClusterDownException. 

2937 """ 

2938 if not stack: 

2939 return [] 

2940 retry_attempts = self._pipe.retry.get_retries() 

2941 while True: 

2942 try: 

2943 return self._send_cluster_commands( 

2944 stack, 

2945 raise_on_error=raise_on_error, 

2946 allow_redirections=allow_redirections, 

2947 ) 

2948 except RedisCluster.ERRORS_ALLOW_RETRY as e: 

2949 if retry_attempts > 0: 

2950 # Try again with the new cluster setup. All other errors 

2951 # should be raised. 

2952 retry_attempts -= 1 

2953 pass 

2954 else: 

2955 raise e 

2956 

2957 def _send_cluster_commands( 

2958 self, stack, raise_on_error=True, allow_redirections=True 

2959 ): 

2960 """ 

2961 Send a bunch of cluster commands to the redis cluster. 

2962 

2963 `allow_redirections` If the pipeline should follow 

2964 `ASK` & `MOVED` responses automatically. If set 

2965 to false it will raise RedisClusterException. 

2966 """ 

2967 # the first time sending the commands we send all of 

2968 # the commands that were queued up. 

2969 # if we have to run through it again, we only retry 

2970 # the commands that failed. 

2971 attempt = sorted(stack, key=lambda x: x.position) 

2972 is_default_node = False 

2973 # build a list of node objects based on node names we need to 

2974 nodes = {} 

2975 

2976 # as we move through each command that still needs to be processed, 

2977 # we figure out the slot number that command maps to, then from 

2978 # the slot determine the node. 

2979 for c in attempt: 

2980 command_policies = self._pipe._policy_resolver.resolve(c.args[0].lower()) 

2981 

2982 while True: 

2983 # refer to our internal node -> slot table that 

2984 # tells us where a given command should route to. 

2985 # (it might be possible we have a cached node that no longer 

2986 # exists in the cluster, which is why we do this in a loop) 

2987 passed_targets = c.options.pop("target_nodes", None) 

2988 if passed_targets and not self._is_nodes_flag(passed_targets): 

2989 target_nodes = self._parse_target_nodes(passed_targets) 

2990 

2991 if not command_policies: 

2992 command_policies = CommandPolicies() 

2993 else: 

2994 if not command_policies: 

2995 command = c.args[0].upper() 

2996 if ( 

2997 len(c.args) >= 2 

2998 and f"{c.args[0]} {c.args[1]}".upper() 

2999 in self._pipe.command_flags 

3000 ): 

3001 command = f"{c.args[0]} {c.args[1]}".upper() 

3002 

3003 # We only could resolve key properties if command is not 

3004 # in a list of pre-defined request policies 

3005 command_flag = self.command_flags.get(command) 

3006 if not command_flag: 

3007 # Fallback to default policy 

3008 if not self._pipe.get_default_node(): 

3009 keys = None 

3010 else: 

3011 keys = self._pipe._get_command_keys(*c.args) 

3012 if not keys or len(keys) == 0: 

3013 command_policies = CommandPolicies() 

3014 else: 

3015 command_policies = CommandPolicies( 

3016 request_policy=RequestPolicy.DEFAULT_KEYED, 

3017 response_policy=ResponsePolicy.DEFAULT_KEYED, 

3018 ) 

3019 else: 

3020 if command_flag in self._pipe._command_flags_mapping: 

3021 command_policies = CommandPolicies( 

3022 request_policy=self._pipe._command_flags_mapping[ 

3023 command_flag 

3024 ] 

3025 ) 

3026 else: 

3027 command_policies = CommandPolicies() 

3028 

3029 target_nodes = self._determine_nodes( 

3030 *c.args, 

3031 request_policy=command_policies.request_policy, 

3032 node_flag=passed_targets, 

3033 ) 

3034 if not target_nodes: 

3035 raise RedisClusterException( 

3036 f"No targets were found to execute {c.args} command on" 

3037 ) 

3038 c.command_policies = command_policies 

3039 if len(target_nodes) > 1: 

3040 raise RedisClusterException( 

3041 f"Too many targets for command {c.args}" 

3042 ) 

3043 

3044 node = target_nodes[0] 

3045 if node == self._pipe.get_default_node(): 

3046 is_default_node = True 

3047 

3048 # now that we know the name of the node 

3049 # ( it's just a string in the form of host:port ) 

3050 # we can build a list of commands for each node. 

3051 node_name = node.name 

3052 if node_name not in nodes: 

3053 redis_node = self._pipe.get_redis_connection(node) 

3054 try: 

3055 connection = get_connection(redis_node) 

3056 except (ConnectionError, TimeoutError): 

3057 for n in nodes.values(): 

3058 n.connection_pool.release(n.connection) 

3059 # Connection retries are being handled in the node's 

3060 # Retry object. Reinitialize the node -> slot table. 

3061 self._nodes_manager.initialize() 

3062 if is_default_node: 

3063 self._pipe.replace_default_node() 

3064 raise 

3065 nodes[node_name] = NodeCommands( 

3066 redis_node.parse_response, 

3067 redis_node.connection_pool, 

3068 connection, 

3069 ) 

3070 nodes[node_name].append(c) 

3071 break 

3072 

3073 # send the commands in sequence. 

3074 # we write to all the open sockets for each node first, 

3075 # before reading anything 

3076 # this allows us to flush all the requests out across the 

3077 # network 

3078 # so that we can read them from different sockets as they come back. 

3079 # we dont' multiplex on the sockets as they come available, 

3080 # but that shouldn't make too much difference. 

3081 try: 

3082 node_commands = nodes.values() 

3083 for n in node_commands: 

3084 n.write() 

3085 

3086 for n in node_commands: 

3087 n.read() 

3088 finally: 

3089 # release all of the redis connections we allocated earlier 

3090 # back into the connection pool. 

3091 # we used to do this step as part of a try/finally block, 

3092 # but it is really dangerous to 

3093 # release connections back into the pool if for some 

3094 # reason the socket has data still left in it 

3095 # from a previous operation. The write and 

3096 # read operations already have try/catch around them for 

3097 # all known types of errors including connection 

3098 # and socket level errors. 

3099 # So if we hit an exception, something really bad 

3100 # happened and putting any oF 

3101 # these connections back into the pool is a very bad idea. 

3102 # the socket might have unread buffer still sitting in it, 

3103 # and then the next time we read from it we pass the 

3104 # buffered result back from a previous command and 

3105 # every single request after to that connection will always get 

3106 # a mismatched result. 

3107 for n in nodes.values(): 

3108 n.connection_pool.release(n.connection) 

3109 

3110 # if the response isn't an exception it is a 

3111 # valid response from the node 

3112 # we're all done with that command, YAY! 

3113 # if we have more commands to attempt, we've run into problems. 

3114 # collect all the commands we are allowed to retry. 

3115 # (MOVED, ASK, or connection errors or timeout errors) 

3116 attempt = sorted( 

3117 ( 

3118 c 

3119 for c in attempt 

3120 if isinstance(c.result, ClusterPipeline.ERRORS_ALLOW_RETRY) 

3121 ), 

3122 key=lambda x: x.position, 

3123 ) 

3124 if attempt and allow_redirections: 

3125 # RETRY MAGIC HAPPENS HERE! 

3126 # send these remaining commands one at a time using `execute_command` 

3127 # in the main client. This keeps our retry logic 

3128 # in one place mostly, 

3129 # and allows us to be more confident in correctness of behavior. 

3130 # at this point any speed gains from pipelining have been lost 

3131 # anyway, so we might as well make the best 

3132 # attempt to get the correct behavior. 

3133 # 

3134 # The client command will handle retries for each 

3135 # individual command sequentially as we pass each 

3136 # one into `execute_command`. Any exceptions 

3137 # that bubble out should only appear once all 

3138 # retries have been exhausted. 

3139 # 

3140 # If a lot of commands have failed, we'll be setting the 

3141 # flag to rebuild the slots table from scratch. 

3142 # So MOVED errors should correct themselves fairly quickly. 

3143 self._pipe.reinitialize_counter += 1 

3144 if self._pipe._should_reinitialized(): 

3145 self._nodes_manager.initialize() 

3146 if is_default_node: 

3147 self._pipe.replace_default_node() 

3148 for c in attempt: 

3149 try: 

3150 # send each command individually like we 

3151 # do in the main client. 

3152 c.result = self._pipe.parent_execute_command(*c.args, **c.options) 

3153 except RedisError as e: 

3154 c.result = e 

3155 

3156 # turn the response back into a simple flat array that corresponds 

3157 # to the sequence of commands issued in the stack in pipeline.execute() 

3158 response = [] 

3159 for c in sorted(stack, key=lambda x: x.position): 

3160 if c.args[0] in self._pipe.cluster_response_callbacks: 

3161 # Remove keys entry, it needs only for cache. 

3162 c.options.pop("keys", None) 

3163 c.result = self._pipe._policies_callback_mapping[ 

3164 c.command_policies.response_policy 

3165 ]( 

3166 self._pipe.cluster_response_callbacks[c.args[0]]( 

3167 c.result, **c.options 

3168 ) 

3169 ) 

3170 response.append(c.result) 

3171 

3172 if raise_on_error: 

3173 self._raise_first_error(stack) 

3174 

3175 return response 

3176 

3177 def _is_nodes_flag(self, target_nodes): 

3178 return isinstance(target_nodes, str) and target_nodes in self._pipe.node_flags 

3179 

3180 def _parse_target_nodes(self, target_nodes): 

3181 if isinstance(target_nodes, list): 

3182 nodes = target_nodes 

3183 elif isinstance(target_nodes, ClusterNode): 

3184 # Supports passing a single ClusterNode as a variable 

3185 nodes = [target_nodes] 

3186 elif isinstance(target_nodes, dict): 

3187 # Supports dictionaries of the format {node_name: node}. 

3188 # It enables to execute commands with multi nodes as follows: 

3189 # rc.cluster_save_config(rc.get_primaries()) 

3190 nodes = target_nodes.values() 

3191 else: 

3192 raise TypeError( 

3193 "target_nodes type can be one of the following: " 

3194 "node_flag (PRIMARIES, REPLICAS, RANDOM, ALL_NODES)," 

3195 "ClusterNode, list<ClusterNode>, or dict<any, ClusterNode>. " 

3196 f"The passed type is {type(target_nodes)}" 

3197 ) 

3198 return nodes 

3199 

3200 def _determine_nodes( 

3201 self, *args, request_policy: RequestPolicy, **kwargs 

3202 ) -> List["ClusterNode"]: 

3203 # Determine which nodes should be executed the command on. 

3204 # Returns a list of target nodes. 

3205 command = args[0].upper() 

3206 if ( 

3207 len(args) >= 2 

3208 and f"{args[0]} {args[1]}".upper() in self._pipe.command_flags 

3209 ): 

3210 command = f"{args[0]} {args[1]}".upper() 

3211 

3212 nodes_flag = kwargs.pop("nodes_flag", None) 

3213 if nodes_flag is not None: 

3214 # nodes flag passed by the user 

3215 command_flag = nodes_flag 

3216 else: 

3217 # get the nodes group for this command if it was predefined 

3218 command_flag = self._pipe.command_flags.get(command) 

3219 

3220 if command_flag in self._pipe._command_flags_mapping: 

3221 request_policy = self._pipe._command_flags_mapping[command_flag] 

3222 

3223 policy_callback = self._pipe._policies_callback_mapping[request_policy] 

3224 

3225 if request_policy == RequestPolicy.DEFAULT_KEYED: 

3226 nodes = policy_callback(command, *args) 

3227 elif request_policy == RequestPolicy.MULTI_SHARD: 

3228 nodes = policy_callback(*args, **kwargs) 

3229 elif request_policy == RequestPolicy.DEFAULT_KEYLESS: 

3230 nodes = policy_callback(args[0]) 

3231 else: 

3232 nodes = policy_callback() 

3233 

3234 if args[0].lower() == "ft.aggregate": 

3235 self._aggregate_nodes = nodes 

3236 

3237 return nodes 

3238 

3239 def multi(self): 

3240 raise RedisClusterException( 

3241 "method multi() is not supported outside of transactional context" 

3242 ) 

3243 

3244 def discard(self): 

3245 raise RedisClusterException( 

3246 "method discard() is not supported outside of transactional context" 

3247 ) 

3248 

3249 def watch(self, *names): 

3250 raise RedisClusterException( 

3251 "method watch() is not supported outside of transactional context" 

3252 ) 

3253 

3254 def unwatch(self, *names): 

3255 raise RedisClusterException( 

3256 "method unwatch() is not supported outside of transactional context" 

3257 ) 

3258 

3259 def delete(self, *names): 

3260 if len(names) != 1: 

3261 raise RedisClusterException( 

3262 "deleting multiple keys is not implemented in pipeline command" 

3263 ) 

3264 

3265 return self.execute_command("DEL", names[0]) 

3266 

3267 def unlink(self, *names): 

3268 if len(names) != 1: 

3269 raise RedisClusterException( 

3270 "unlinking multiple keys is not implemented in pipeline command" 

3271 ) 

3272 

3273 return self.execute_command("UNLINK", names[0]) 

3274 

3275 

3276class TransactionStrategy(AbstractStrategy): 

3277 NO_SLOTS_COMMANDS = {"UNWATCH"} 

3278 IMMEDIATE_EXECUTE_COMMANDS = {"WATCH", "UNWATCH"} 

3279 UNWATCH_COMMANDS = {"DISCARD", "EXEC", "UNWATCH"} 

3280 SLOT_REDIRECT_ERRORS = (AskError, MovedError) 

3281 CONNECTION_ERRORS = ( 

3282 ConnectionError, 

3283 OSError, 

3284 ClusterDownError, 

3285 SlotNotCoveredError, 

3286 ) 

3287 

3288 def __init__(self, pipe: ClusterPipeline): 

3289 super().__init__(pipe) 

3290 self._explicit_transaction = False 

3291 self._watching = False 

3292 self._pipeline_slots: Set[int] = set() 

3293 self._transaction_connection: Optional[Connection] = None 

3294 self._executing = False 

3295 self._retry = copy(self._pipe.retry) 

3296 self._retry.update_supported_errors( 

3297 RedisCluster.ERRORS_ALLOW_RETRY + self.SLOT_REDIRECT_ERRORS 

3298 ) 

3299 

3300 def _get_client_and_connection_for_transaction(self) -> Tuple[Redis, Connection]: 

3301 """ 

3302 Find a connection for a pipeline transaction. 

3303 

3304 For running an atomic transaction, watch keys ensure that contents have not been 

3305 altered as long as the watch commands for those keys were sent over the same 

3306 connection. So once we start watching a key, we fetch a connection to the 

3307 node that owns that slot and reuse it. 

3308 """ 

3309 if not self._pipeline_slots: 

3310 raise RedisClusterException( 

3311 "At least a command with a key is needed to identify a node" 

3312 ) 

3313 

3314 node: ClusterNode = self._nodes_manager.get_node_from_slot( 

3315 list(self._pipeline_slots)[0], False 

3316 ) 

3317 redis_node: Redis = self._pipe.get_redis_connection(node) 

3318 if self._transaction_connection: 

3319 if not redis_node.connection_pool.owns_connection( 

3320 self._transaction_connection 

3321 ): 

3322 previous_node = self._nodes_manager.find_connection_owner( 

3323 self._transaction_connection 

3324 ) 

3325 previous_node.connection_pool.release(self._transaction_connection) 

3326 self._transaction_connection = None 

3327 

3328 if not self._transaction_connection: 

3329 self._transaction_connection = get_connection(redis_node) 

3330 

3331 return redis_node, self._transaction_connection 

3332 

3333 def execute_command(self, *args, **kwargs): 

3334 slot_number: Optional[int] = None 

3335 if args[0] not in ClusterPipeline.NO_SLOTS_COMMANDS: 

3336 slot_number = self._pipe.determine_slot(*args) 

3337 

3338 if ( 

3339 self._watching or args[0] in self.IMMEDIATE_EXECUTE_COMMANDS 

3340 ) and not self._explicit_transaction: 

3341 if args[0] == "WATCH": 

3342 self._validate_watch() 

3343 

3344 if slot_number is not None: 

3345 if self._pipeline_slots and slot_number not in self._pipeline_slots: 

3346 raise CrossSlotTransactionError( 

3347 "Cannot watch or send commands on different slots" 

3348 ) 

3349 

3350 self._pipeline_slots.add(slot_number) 

3351 elif args[0] not in self.NO_SLOTS_COMMANDS: 

3352 raise RedisClusterException( 

3353 f"Cannot identify slot number for command: {args[0]}," 

3354 "it cannot be triggered in a transaction" 

3355 ) 

3356 

3357 return self._immediate_execute_command(*args, **kwargs) 

3358 else: 

3359 if slot_number is not None: 

3360 self._pipeline_slots.add(slot_number) 

3361 

3362 return self.pipeline_execute_command(*args, **kwargs) 

3363 

3364 def _validate_watch(self): 

3365 if self._explicit_transaction: 

3366 raise RedisError("Cannot issue a WATCH after a MULTI") 

3367 

3368 self._watching = True 

3369 

3370 def _immediate_execute_command(self, *args, **options): 

3371 return self._retry.call_with_retry( 

3372 lambda: self._get_connection_and_send_command(*args, **options), 

3373 self._reinitialize_on_error, 

3374 ) 

3375 

3376 def _get_connection_and_send_command(self, *args, **options): 

3377 redis_node, connection = self._get_client_and_connection_for_transaction() 

3378 return self._send_command_parse_response( 

3379 connection, redis_node, args[0], *args, **options 

3380 ) 

3381 

3382 def _send_command_parse_response( 

3383 self, conn, redis_node: Redis, command_name, *args, **options 

3384 ): 

3385 """ 

3386 Send a command and parse the response 

3387 """ 

3388 

3389 conn.send_command(*args) 

3390 output = redis_node.parse_response(conn, command_name, **options) 

3391 

3392 if command_name in self.UNWATCH_COMMANDS: 

3393 self._watching = False 

3394 return output 

3395 

3396 def _reinitialize_on_error(self, error): 

3397 if self._watching: 

3398 if type(error) in self.SLOT_REDIRECT_ERRORS and self._executing: 

3399 raise WatchError("Slot rebalancing occurred while watching keys") 

3400 

3401 if ( 

3402 type(error) in self.SLOT_REDIRECT_ERRORS 

3403 or type(error) in self.CONNECTION_ERRORS 

3404 ): 

3405 if self._transaction_connection: 

3406 self._transaction_connection = None 

3407 

3408 self._pipe.reinitialize_counter += 1 

3409 if self._pipe._should_reinitialized(): 

3410 self._nodes_manager.initialize() 

3411 self.reinitialize_counter = 0 

3412 else: 

3413 if isinstance(error, AskError): 

3414 self._nodes_manager.update_moved_exception(error) 

3415 

3416 self._executing = False 

3417 

3418 def _raise_first_error(self, responses, stack): 

3419 """ 

3420 Raise the first exception on the stack 

3421 """ 

3422 for r, cmd in zip(responses, stack): 

3423 if isinstance(r, Exception): 

3424 self.annotate_exception(r, cmd.position + 1, cmd.args) 

3425 raise r 

3426 

3427 def execute(self, raise_on_error: bool = True) -> List[Any]: 

3428 stack = self._command_queue 

3429 if not stack and (not self._watching or not self._pipeline_slots): 

3430 return [] 

3431 

3432 return self._execute_transaction_with_retries(stack, raise_on_error) 

3433 

3434 def _execute_transaction_with_retries( 

3435 self, stack: List["PipelineCommand"], raise_on_error: bool 

3436 ): 

3437 return self._retry.call_with_retry( 

3438 lambda: self._execute_transaction(stack, raise_on_error), 

3439 self._reinitialize_on_error, 

3440 ) 

3441 

3442 def _execute_transaction( 

3443 self, stack: List["PipelineCommand"], raise_on_error: bool 

3444 ): 

3445 if len(self._pipeline_slots) > 1: 

3446 raise CrossSlotTransactionError( 

3447 "All keys involved in a cluster transaction must map to the same slot" 

3448 ) 

3449 

3450 self._executing = True 

3451 

3452 redis_node, connection = self._get_client_and_connection_for_transaction() 

3453 

3454 stack = chain( 

3455 [PipelineCommand(("MULTI",))], 

3456 stack, 

3457 [PipelineCommand(("EXEC",))], 

3458 ) 

3459 commands = [c.args for c in stack if EMPTY_RESPONSE not in c.options] 

3460 packed_commands = connection.pack_commands(commands) 

3461 connection.send_packed_command(packed_commands) 

3462 errors = [] 

3463 

3464 # parse off the response for MULTI 

3465 # NOTE: we need to handle ResponseErrors here and continue 

3466 # so that we read all the additional command messages from 

3467 # the socket 

3468 try: 

3469 redis_node.parse_response(connection, "MULTI") 

3470 except ResponseError as e: 

3471 self.annotate_exception(e, 0, "MULTI") 

3472 errors.append(e) 

3473 except self.CONNECTION_ERRORS as cluster_error: 

3474 self.annotate_exception(cluster_error, 0, "MULTI") 

3475 raise 

3476 

3477 # and all the other commands 

3478 for i, command in enumerate(self._command_queue): 

3479 if EMPTY_RESPONSE in command.options: 

3480 errors.append((i, command.options[EMPTY_RESPONSE])) 

3481 else: 

3482 try: 

3483 _ = redis_node.parse_response(connection, "_") 

3484 except self.SLOT_REDIRECT_ERRORS as slot_error: 

3485 self.annotate_exception(slot_error, i + 1, command.args) 

3486 errors.append(slot_error) 

3487 except self.CONNECTION_ERRORS as cluster_error: 

3488 self.annotate_exception(cluster_error, i + 1, command.args) 

3489 raise 

3490 except ResponseError as e: 

3491 self.annotate_exception(e, i + 1, command.args) 

3492 errors.append(e) 

3493 

3494 response = None 

3495 # parse the EXEC. 

3496 try: 

3497 response = redis_node.parse_response(connection, "EXEC") 

3498 except ExecAbortError: 

3499 if errors: 

3500 raise errors[0] 

3501 raise 

3502 

3503 self._executing = False 

3504 

3505 # EXEC clears any watched keys 

3506 self._watching = False 

3507 

3508 if response is None: 

3509 raise WatchError("Watched variable changed.") 

3510 

3511 # put any parse errors into the response 

3512 for i, e in errors: 

3513 response.insert(i, e) 

3514 

3515 if len(response) != len(self._command_queue): 

3516 raise InvalidPipelineStack( 

3517 "Unexpected response length for cluster pipeline EXEC." 

3518 " Command stack was {} but response had length {}".format( 

3519 [c.args[0] for c in self._command_queue], len(response) 

3520 ) 

3521 ) 

3522 

3523 # find any errors in the response and raise if necessary 

3524 if raise_on_error or len(errors) > 0: 

3525 self._raise_first_error( 

3526 response, 

3527 self._command_queue, 

3528 ) 

3529 

3530 # We have to run response callbacks manually 

3531 data = [] 

3532 for r, cmd in zip(response, self._command_queue): 

3533 if not isinstance(r, Exception): 

3534 command_name = cmd.args[0] 

3535 if command_name in self._pipe.cluster_response_callbacks: 

3536 r = self._pipe.cluster_response_callbacks[command_name]( 

3537 r, **cmd.options 

3538 ) 

3539 data.append(r) 

3540 return data 

3541 

3542 def reset(self): 

3543 self._command_queue = [] 

3544 

3545 # make sure to reset the connection state in the event that we were 

3546 # watching something 

3547 if self._transaction_connection: 

3548 try: 

3549 if self._watching: 

3550 # call this manually since our unwatch or 

3551 # immediate_execute_command methods can call reset() 

3552 self._transaction_connection.send_command("UNWATCH") 

3553 self._transaction_connection.read_response() 

3554 # we can safely return the connection to the pool here since we're 

3555 # sure we're no longer WATCHing anything 

3556 node = self._nodes_manager.find_connection_owner( 

3557 self._transaction_connection 

3558 ) 

3559 node.redis_connection.connection_pool.release( 

3560 self._transaction_connection 

3561 ) 

3562 self._transaction_connection = None 

3563 except self.CONNECTION_ERRORS: 

3564 # disconnect will also remove any previous WATCHes 

3565 if self._transaction_connection: 

3566 self._transaction_connection.disconnect() 

3567 

3568 # clean up the other instance attributes 

3569 self._watching = False 

3570 self._explicit_transaction = False 

3571 self._pipeline_slots = set() 

3572 self._executing = False 

3573 

3574 def send_cluster_commands( 

3575 self, stack, raise_on_error=True, allow_redirections=True 

3576 ): 

3577 raise NotImplementedError( 

3578 "send_cluster_commands cannot be executed in transactional context." 

3579 ) 

3580 

3581 def multi(self): 

3582 if self._explicit_transaction: 

3583 raise RedisError("Cannot issue nested calls to MULTI") 

3584 if self._command_queue: 

3585 raise RedisError( 

3586 "Commands without an initial WATCH have already been issued" 

3587 ) 

3588 self._explicit_transaction = True 

3589 

3590 def watch(self, *names): 

3591 if self._explicit_transaction: 

3592 raise RedisError("Cannot issue a WATCH after a MULTI") 

3593 

3594 return self.execute_command("WATCH", *names) 

3595 

3596 def unwatch(self): 

3597 if self._watching: 

3598 return self.execute_command("UNWATCH") 

3599 

3600 return True 

3601 

3602 def discard(self): 

3603 self.reset() 

3604 

3605 def delete(self, *names): 

3606 return self.execute_command("DEL", *names) 

3607 

3608 def unlink(self, *names): 

3609 return self.execute_command("UNLINK", *names)