Enhance Registry to Remember Blacklisted Servers for Actors

[aratz-lasa]

Description

This issue is to enhance the registry functionality by implementing the ability to remember blacklisted servers for actors. Currently, the registry only knows the single blacklisted server ID provided by the caller. However, it would be beneficial to accumulate a set of known blacklisted servers for each actor, along with their corresponding blacklist end times.

Enhancement Proposal

1. Modify the registry implementation to maintain a blacklist set for each actor, storing the server IDs of the blacklisted servers.
2. Include an associated blacklist end time for each blacklisted server in order to track the duration of the blacklist.
3. Update the registry methods to handle the accumulation and management of blacklisted servers for actors, considering factors such as expiration of blacklists and proper handling of subsequent requests.
4. Implement appropriate data structures and algorithms to efficiently store and retrieve blacklisted servers for actors.
5. Consider the impact on performance and resource utilization while designing and implementing the enhanced blacklist functionality.
6. Write tests to validate the behavior and correctness of the registry with the new blacklisted server tracking feature.
7. Evaluate the effectiveness of the enhanced registry in real-world scenarios and gather feedback for further improvements.

Additional information

The ability to remember blacklisted servers for actors will provide several benefits, including:

• Improved fault tolerance: The registry can avoid routing requests to blacklisted servers, ensuring better availability and reliability.
• Enhanced load balancing: The blacklist information can be utilized to distribute actor replicas more evenly across available servers.
• Efficient blacklist management: Actors can be dynamically blacklisted and removed from blacklists based on configurable policies or external signals.
• Fine-grained control: The registry can keep track of blacklisted servers at the actor level, enabling targeted handling and isolation of problematic servers.

This issue will serve as a placeholder to explore and experiment with the proposed enhancements, gather feedback, and conduct further investigations.

[richardartoul]

I added some comments here that I think are relevant: https://github.com/richardartoul/nola/issues/72