deploy.yaml

apiVersion: apps/v1
kind: DaemonSet
metadata:
  labels:
    k8s-app: gpu-memory-monitor
  name: gpu-memory-monitor
  namespace: kube-system
spec:
  selector:
    matchLabels:
      k8s-app: gpu-memory-monitor
  template:
    metadata:
      labels:
        k8s-app: gpu-memory-monitor
    spec:
      affinity:
        nodeAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            nodeSelectorTerms:
              - matchExpressions:
                  - key: gpushare
                    operator: In
                    values:
                      - "true"
      containers:
        - env:
            - name: NVIDIA_VISIBLE_DEVICES
              value: all
            - name: NVIDIA_DRIVER_CAPABILITIES
              value: utility
          image: gpu-memory-monitor:v1
          imagePullPolicy: IfNotPresent
          name: gpu-memory-monitor
          resources:
            requests:
              cpu: 500m
          volumeMounts:
            - mountPath: /var/run
              name: docker-sock
              readOnly: false
      hostNetwork: true
      restartPolicy: Always
      volumes:
        - hostPath:
            path: /var/run
          name: docker-sock