#1335: feat: implement distributed caching layer · .git

feat: implement distributed caching layer #1335

✓ Ready to merge enhancement performance

AX

@alexchen assigned to themself

📊 3 commits

📄 12 changed files

Conversation 4

Commits 3

Checks 6

Files changed 12

Description

Introduces a new distributed caching layer using Redis cluster with automatic failover and TTL-based eviction policies. This replaces the legacy in-memory singleton cache that was causing memory leaks under high load.

Key changes:

Connection pooling with circuit breaker pattern
Async serialization/deserialization using msgpack
Automatic shard discovery & rebalancing
Drop-in replacement for CacheManager

const cache = new DistributedCache({
  nodes: process.env.REDIS_NODES?.split(',') || [],
  ttl: 3600,
  serializer: 'msgpack'
});

Benchmark results show a 3.2x throughput increase and 40% reduction in p99 latency for cached endpoints. Ready for review & merge into develop.

MR

marcus_lee reviewed 2 hours ago

The circuit breaker implementation looks solid. One suggestion: consider adding a configuration flag to disable cluster mode for local dev environments to avoid pulling up a full Redis stack during testing.

src/cache/client.ts +3 -1

- export const CLUSTER_ENABLED = true;

+ export const CLUSTER_ENABLED = process.env.CACHE_CLUSTER_MODE === 'true';

AX

alexchen author commented 1 hour ago

@marcus_lee Agreed. I've updated the config loader to fallback to standalone mode when REDIS_NODES is undefined or empty. Also added a Dockerfile.cache-dev for easy local testing. Pushed to feat/cache-v2.

SJ

sarah_jenkins reviewed 45 minutes ago

LGTM! The serialization fallback to JSON is a nice safety net. Approving.