[grpc] Support gRPC standard health check

[CatherineSue]

Motivation

This PR adds a HealthCheck servicer following gRPC's standard Health Checking Protocol, grpc.health.v1.Health

This enables native Kubernetes gRPC health probes without requiring custom scripts or tools.

Reference: https://kubernetes.io/docs/tasks/configure-pod-container/configure-liveness-readiness-startup-probes/#define-a-grpc-liveness-probe

Health Check Services

The gRPC server exposes two health check services:

1. Overall Server Health (service: "")

   • Checks if the gRPC server process is running
   • Fast, lightweight check
   • Recommended for liveness probes

2. SGLang Service Health (service: "sglang.grpc.scheduler.SglangScheduler")

   • Checks if the model is loaded and ready to serve
   • Checks if scheduler is responsive
   • Recommended for readiness probes

Modifications

• Add grpcio-health-checking into pyproject.toml. This is required for HealthServicer
  • NOTE: using customized exec would require grpcurl dependency, and it would be slower, so it has no advantage over the current approach
• Add SGLangHealthServicer to inherit from standard health_pb2_grpc.HealthServicer
• Move _run_scheduler_with_signal_handling and _launch_scheduler_process_only to grpc/scheduler_handler.py as grpc_server.py is getting too long

Kubernetes Configuration

Basic Example

apiVersion: v1
kind: Pod
metadata:
  name: sglang-server
spec:
  containers:
  - name: sglang
    image: your-sglang-image:latest
    ports:
    - containerPort: 30000
      name: grpc

    # Liveness: Restart if process dies
    livenessProbe:
      grpc:
        port: 30000
        service: ""  # Overall server health
      initialDelaySeconds: 60
      periodSeconds: 10
      timeoutSeconds: 5
      failureThreshold: 3

    # Readiness: Don't send traffic until ready
    readinessProbe:
      grpc:
        port: 30000
        service: "sglang.grpc.scheduler.SglangScheduler"
      initialDelaySeconds: 30
      periodSeconds: 5
      timeoutSeconds: 3
      failureThreshold: 3

Benchmarking and Profiling

Checklist

• Format your code according to the Format code with pre-commit.
• Add unit tests according to the Run and add unit tests.
• Update documentation according to Write documentations.
• Provide accuracy and speed benchmark results according to Test the accuracy and Benchmark the speed.

[gemini-code-assist]

Warning

You have reached your daily quota limit. Please wait up to 24 hours and I will start processing your requests again!

[slin1237]

tested manually