3e03879f68
Co-authored-by: Kangyan Zhou <kangyan.zhou@radixark.ai> Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
91 lines
2.8 KiB
YAML
91 lines
2.8 KiB
YAML
apiVersion: apps/v1
|
|
kind: Deployment
|
|
metadata:
|
|
name: sgl-router
|
|
namespace: sgl-router-test
|
|
spec:
|
|
replicas: 1
|
|
selector:
|
|
matchLabels:
|
|
app: sgl-router
|
|
template:
|
|
metadata:
|
|
labels:
|
|
app: sgl-router
|
|
spec:
|
|
serviceAccountName: sgl-router
|
|
# Must exceed --shutdown-drain-secs plus the time in-flight requests
|
|
# need after the pause, or the pod is SIGKILLed mid-drain and the drain
|
|
# has bought nothing.
|
|
terminationGracePeriodSeconds: 40
|
|
containers:
|
|
- name: router
|
|
image: sgl-router:e2e
|
|
imagePullPolicy: Never
|
|
# Configured entirely via CLI flags. K8s EndpointSlice discovery
|
|
# watches `app=sglang` pods in this namespace. The aggressive
|
|
# circuit breaker (threshold 1, 5s cool-down) lets a terminating
|
|
# pod's connection-refused immediately drop it from the candidate
|
|
# set — the reconciliation tests scale workers rapidly and depend
|
|
# on fast eviction to absorb the churn.
|
|
args:
|
|
- "--host"
|
|
- "0.0.0.0"
|
|
- "--port"
|
|
- "8090"
|
|
- "--model-id"
|
|
- "tiny"
|
|
- "--tokenizer-path"
|
|
- "/etc/tokenizer/tiny.json"
|
|
- "--policy"
|
|
- "round_robin"
|
|
- "--cb-threshold"
|
|
- "1"
|
|
- "--cb-cool-down-secs"
|
|
- "5"
|
|
- "--service-discovery"
|
|
- "--service-discovery-namespace"
|
|
- "sgl-router-test"
|
|
- "--selector"
|
|
- "app=sglang"
|
|
# On SIGTERM, keep serving with /readyz at 503 so the endpoint
|
|
# removal reaches kube-proxy before the listener closes. Sized for
|
|
# the deletionTimestamp path, which does not wait on a probe.
|
|
# Probe-driven deregistration would instead need this above the
|
|
# readinessProbe's failureThreshold * periodSeconds below.
|
|
# test_shutdown_drain.py reads this value; do not restate it there.
|
|
- "--shutdown-drain-secs"
|
|
- "8"
|
|
# Declared so the startup advisory compares the drain against this
|
|
# pod's real grace period instead of assuming the k8s default.
|
|
- "--termination-grace-secs"
|
|
- "40"
|
|
ports:
|
|
- containerPort: 8090
|
|
name: http
|
|
readinessProbe:
|
|
httpGet:
|
|
path: /readyz
|
|
port: 8090
|
|
initialDelaySeconds: 3
|
|
periodSeconds: 3
|
|
livenessProbe:
|
|
httpGet:
|
|
path: /healthz
|
|
port: 8090
|
|
initialDelaySeconds: 5
|
|
periodSeconds: 10
|
|
---
|
|
apiVersion: v1
|
|
kind: Service
|
|
metadata:
|
|
name: sgl-router
|
|
namespace: sgl-router-test
|
|
spec:
|
|
selector:
|
|
app: sgl-router
|
|
ports:
|
|
- name: http
|
|
port: 8090
|
|
targetPort: 8090
|