apiVersion: apps/v1 kind: Deployment metadata: name: deepseek-ultrafast-redis spec: replicas: 1 selector: matchLabels: {app: deepseek-ultrafast-redis} template: metadata: labels: {app: deepseek-ultrafast-redis} spec: containers: - name: redis image: redis:7 args: ["--save", "", "--appendonly", "no"] ports: - containerPort: 6379 --- apiVersion: v1 kind: Service metadata: name: deepseek-ultrafast-redis spec: selector: {app: deepseek-ultrafast-redis} ports: - port: 6379 --- apiVersion: apps/v1 kind: Deployment metadata: name: deepseek-ultrafast spec: replicas: 2 selector: matchLabels: {app: deepseek-ultrafast} template: metadata: labels: {app: deepseek-ultrafast} spec: hostNetwork: true dnsPolicy: ClusterFirstWithHostNet terminationGracePeriodSeconds: 60 containers: - name: serve image: deepseek-ultrafast:latest args: ["--gpus", "0,1,2,3", "--port", "8000"] env: - name: DSUF_KV_HOST valueFrom: fieldRef: {fieldPath: status.podIP} - name: DSUF_REDIS_URL value: redis://deepseek-ultrafast-redis:6379 ports: - {name: http, containerPort: 8000} - {name: kv, containerPort: 19500} resources: limits: nvidia.com/gpu: 4 rdma/hca: 1 startupProbe: httpGet: {path: /readyz, port: http} periodSeconds: 5 failureThreshold: 120 readinessProbe: httpGet: {path: /readyz, port: http} periodSeconds: 2 volumeMounts: - {name: models, mountPath: /models, readOnly: true} volumes: - name: models persistentVolumeClaim: claimName: deepseek-ultrafast-models --- apiVersion: v1 kind: Service metadata: name: deepseek-ultrafast spec: selector: {app: deepseek-ultrafast} ports: - {name: http, port: 8000, targetPort: http}