Folding @ Home GPU Kubernetes Deployment

folding-at-home-deployment.yaml

apiVersion: apps/v1
kind: Deployment
metadata:
  name: folding-at-home
spec:
  progressDeadlineSeconds: 25920000
  replicas: 1
  revisionHistoryLimit: 2
  selector:
    matchLabels:
      app.kubernetes.io/name: folding-at-home
  strategy:
    type: RollingUpdate
    rollingUpdate:
      maxUnavailable: "50%"
      maxSurge: 0
  template:
    metadata:
      annotations:
        qos.coreweave.cloud/drain: "false"
      labels:
        app.kubernetes.io/name: folding-at-home
    spec:
      - name: folding
        image: coreweave/fah-gpu:7.5.1
        imagePullPolicy: IfNotPresent
        #command: ['bash', '-c', 'FAHClient --user=CoreWeave --team=236584 --gpu=true --smp=true --priority=low --cpu-usage=90 --web-allow=0/0 --allow=0/0 --gpu-usage=100']
        ports:
        - containerPort: 7396
          name: ui
          protocol: TCP
        livenessProbe:
          failureThreshold: 3
          initialDelaySeconds: 30
          periodSeconds: 10
          successThreshold: 1
          tcpSocket:
            port: 7396
          timeoutSeconds: 1
        readinessProbe:
          failureThreshold: 3
          initialDelaySeconds: 5
          periodSeconds: 10
          successThreshold: 1
          tcpSocket:
            port: 7396
          timeoutSeconds: 1
        resources:
          limits:
            cpu: "3"
            memory: 8Gi
            nvidia.com/gpu: 8 # Allocate 8 GPUs
          requests:
            cpu: 100m
            memory: 5Gi
        volumeMounts:
        - name: userdir
          mountPath: /root
      volumes:
      - name: userdir
        hostPath:
          path: "/mnt/local/fah"
          type: DirectoryOrCreate
      automountServiceAccountToken: false
      restartPolicy: Always
      terminationGracePeriodSeconds: 10