Understanding the Search Query — Part III
Docker Deployment
Two containers communicate over a bridge network (172.16.0.0/16):
| Container | Image | Port | IP |
|---|---|---|---|
| Crocodile Model | crocodile-model:latest | 8501 | 172.16.0.2 |
| Service Orchestrator | service | 8080 | 172.16.0.3 |
The orchestrator consumes the Crocodile model’s REST API, applies business logic (query normalization, dimension detection), then returns responses.
Kubernetes Orchestration
Scaled using deployment manifests:
- Replicas: 2 for redundancy
- Resource limits: 2GB memory, 1 CPU per container
- Resource requests: 512MB memory, 0.5 CPU per container
- Persistent Volume: Attached for data retention across pod replicas
Traffic management via Istio Ingress VirtualService for routing and load distribution.
Health monitoring via liveness and readiness probes — orchestrator exposes /healthz endpoint.