CircuitBreaker resiliency strategy over ConnectionMultiplexer applicability questions.

[prokopCerny]

Hello! My team is running a service connecting to Redis (specifically Azure Cache for Redis, with soon plans to migrate to Azure Managed Redis in OSS cluster policy mode), and we're considering adding a CircuitBreaker policy (using Polly) around the ConnectionMultiplexer from StackExchange.Redis object.

We were wondering about the appropriateness and best practices in setting such resiliency strategy up, as our Redises are sharded, and the ConnectionMultiplexer internally multiplexes over them (and ConnectionMultiplexers are Singletons across the same database), which leads to the concerns about breaking the circuit only due to failures of few shards, while the cluster overall might still be working, but also wondering if there are other concerns, edge-cases we hadn't considered how a CB might interact during multiplexing (primaries, replicas, etc.).

We want to add some more resiliency into our solution, and any insights, best practices and suggestions would be welcome. Thank you!