[Repository Pattern] Pin keyspaces to single slot only for index keys on cluster mode

[sureshkmit]

Spring Data Redis repository pattern can confine only index keys within a single Redis slot when operating in cluster mode. Entity keys can be distributed across multiple slots as per the cluster's default hash algorithm. Commands (such as SINTER and SUNION) can still be processed on the server side as all involved index keys map to the same slot. Distributing entity keys across multiple Redis slots in a cluster environment can reduce hot spots and improved scalability and performance.

https://docs.spring.io/spring-data/redis/reference/redis/redis-repositories/cluster.html

Example NOTE: Curly brace only on index keys.

Key	Type	Slot	Node
people:e2c7dcee-b8cd-4424-883e-736ce564363e	id for hash	15171	127.0.0.1:7379
people:a9d4b3a0-50d3-4538-a2fc-f7fc2581ee56	id for hash	7373	127.0.0.1:7379
{people}:firstname:rand	index	2399	127.0.0.1:7379
{people}:firstname:mike	index	2399	127.0.0.1:7379

[christophstrobl]

If you would like us to spend some time helping you to diagnose the problem, please spend some time describing it. Right now, we struggle to understand what you want to achieve.

[sureshkmit]

Our current implementation utilizes the repository pattern to cache millions of objects, with indexing enabled on specific attributes. By confining all keys to a single Redis slot, we're effectively centralizing all data within a single shard. This approach negates the benefits of Redis Cluster mode, as horizontal scaling across multiple shards is hindered. To leverage the full potential of Redis Cluster, the library should distribute the objects data across the cluster. I believe restricting index keys to a single slot is adequate.