 Janis
|
f327441089
|
feat(coordinator): scaffold microservice for <300ms failover and load balancing
- Add coordinator module with gRPC stream-based instance health detection
- Implement InstanceHeartbeatService in core: bidirectional stream to coordinator every 200ms
- Track game subscriptions per core via Redis Sets (SADD/SREM)
- Add gRPC handlers for batch resubscribe/unsubscribe/evict/drain operations
- Implement coordinator services: InstanceRegistry, FailoverService, LoadBalancer, AutoScaler, CacheEvictionManager
- Add REST API for metrics and manual failover/rebalance/scaling
- Proto definition: coordinator_service.proto with HeartbeatStream + batch game operations
- Failover timeline: gRPC stream drop (50-200ms) → game migration (<300ms target)
- Support for Argo Rollouts auto-scaling (k8s CRD patching via Fabric8 client)
Note: Proto compilation issues documented in COORDINATOR_IMPLEMENTATION.md. Requires:
- Add task dependency: tasks.compileScala dependsOn tasks.compileJava
- Fix deprecated @Inject var = _ → = uninitialized syntax
- Implement remaining service methods (gRPC clients, FailoverService distribution)
Co-Authored-By: Claude Haiku 4.5 <noreply@anthropic.com>
|
2026-04-26 08:34:53 +02:00 |
|