GH-10546: Fix RedisLockRegistry for cluster

Fixes: #10546

* pass `unLockChannelKey` as `ARGV[2]` instead of `KEYS[2]` to make this unlock script compatible with both Redis cluster and AWS Elastic valkey cluster
* add documentation for AWS ElastiCache Valkey Support in `RedisLockRegistry`

Signed-off-by: Severin Kistler <kistlerseverin@gmail.com>

**Auto-cherry-pick to `6.5.x` & `6.4.x`**

Author: kistlers

spring-integration-redis/src/main/java/org/springframework/integration/redis/util/RedisLockRegistry.java

@@ -670,7 +670,7 @@ private final class RedisPubSubLock extends RedisLock {
 		private static final String UNLINK_UNLOCK_SCRIPT = """
 				local lockClientId = redis.call('GET', KEYS[1])
 				if (lockClientId == ARGV[1] and redis.call('UNLINK', KEYS[1]) == 1) then
-					redis.call('PUBLISH', KEYS[2], KEYS[1])
+					redis.call('PUBLISH', ARGV[2], KEYS[1])
 					return true
 				end
 				return false
@@ -693,8 +693,8 @@ protected boolean tryRedisLockInner(long time, long expireAfter)
 		protected boolean removeLockKeyInnerUnlink() {
 			final String unLockChannelKey = RedisLockRegistry.this.unLockChannelKey + ":" + this.lockKey;
 			return Boolean.TRUE.equals(RedisLockRegistry.this.redisTemplate.execute(
-					UNLINK_UNLOCK_REDIS_SCRIPT, List.of(this.lockKey, unLockChannelKey),
-					RedisLockRegistry.this.clientId));
+					UNLINK_UNLOCK_REDIS_SCRIPT, List.of(this.lockKey),
+					RedisLockRegistry.this.clientId, unLockChannelKey));
 		}
 
 		private boolean subscribeLock(long time, long expireAfter) throws ExecutionException, InterruptedException {

src/reference/antora/modules/ROOT/pages/redis.adoc

@@ -865,4 +865,27 @@ When it is set, the lock will be automatically renewed every `1/3` of the expira
 
 Starting with version 7.0, the `RedisLock` implements `DistributedLock` interface to support the feature of customized time-to-live (TTL) for the lock status data.
 A `RedisLock` can now be acquired using the `lock(Duration ttl)` or `tryLock(long time, TimeUnit unit, Duration ttl)` method, with a specified time-to-live (TTL) value.
-The `RedisLockRegistry` now provides new `renewLock(Object lockKey, Duration ttl)` method, allowing you to renew the lock with a custom time-to-live value.
+The `RedisLockRegistry` now provides new `renewLock(Object lockKey, Duration ttl)` method, allowing you to renew the lock with a custom time-to-live value.
+
+[[elasticache-valkey-cluster]]
+=== AWS ElastiCache for Valkey Support in cluster mode
+
+Starting with version 6.4.9/6.5.4/7.0.0, `RedisLockRegistry` supports AWS Elasticache for Valkey in cluster mode.
+In this version of valkey (a redis drop-in replacement), all PubSub operations (`PUBLISH`, `SUBSCRIBE`, etc.) use their sharded variants (`SPUBLISH`, `SSUBSCRIBE`, etc.) internally.
+If you are observing errors in the form of:
+
+[source]
+----
+Caused by: io.lettuce.core.RedisCommandExecutionException: ERR Script attempted to access keys that do not hash to the same slot script: b2dedc0ab01c17f9f20e3e6ddb62dcb6afbed0bd, on @user_script:3.
+----
+
+in the `unlock` step of the `RedisLockRegistry`,  you have to supply a lock key that includes a hash tag `{...}` to ensure all operations in the `unlock` script are hashed to the same cluster slot/shard, e.g.:
+
+[source]
+----
+RedisLockRegistry lockRegistry = new RedisLockRegistry("my-lock-key{choose_your_tag}");
+
+lockRegistry.lock();
+# critical section
+lockRegistry.unlock();
+----