job_mpi.yaml

apiVersion: v1
kind: Service
metadata:
  name: blake-container
  namespace: nnf-system
spec:
  clusterIP: None
  selector:
    cray.nnf.container.workflow1: "true"
  # ports:
  #   - protocol: TCP
  #     port: 80
  #     name: web
  #   - protocol: TCP
  #     port: 22
  #     name: ssh
---
# apiVersion: v1
# kind: Secret
# metadata:
#   name: blake-ssh-auth
#   namespace: nnf-system
# type: kubernetes.io/ssh-auth
# data:
#   ssh-privatekey: |
#           MIIEpQIBAAKCAQEAulqb/Y ...
#   ssh-publickey: |
---
apiVersion: batch/v1
kind: Job
metadata:
  name: blake-container2
  namespace: nnf-system
spec:
  # backoffLimit: 0
  # activeDeadlineSeconds: 0
  # ttlSecondsAfterFinished: 10
  # completions: 0
  # parallelism: 1
  template:
    metadata:
      labels:
        cray.nnf.container.workflow1: "true"
    spec:
      # restartPolicy: OnFailure
      restartPolicy: Never
      hostname: blake-container
      subdomain: blake-container
      shareProcessNamespace: true
      containers:
        - name: container
          env:
            - name: OMPI_MCA_orte_keep_fqdn_hostnames
              value: "true"
          image: ghcr.io/nearnodeflash/nnf-mfu:82aa3c4b6655433dacbe3294ad359161c62219c6
          command: 
            - /usr/sbin/sshd
          args:
            - -De
          securityContext:
            privileged: true
            capabilities:
              add: ["SETUID", "SETGID", "MKNOD"]
          volumeMounts:
            - name: ssh-auth
              mountPath: /root/.ssh
            # - name: kube-api-access-c5gl9
            #   mountPath: /var/run/secrets/kubernetes.io/serviceaccount
            #   readOnly: true

      tolerations:
        - effect: NoSchedule
          key: cray.nnf.node
          operator: Equal
          value: "true"
      nodeSelector:
        kubernetes.io/hostname: "kind-worker2"
      volumes:
        - name: ssh-auth
          secret:
            defaultMode: 384
            items:
            - key: ssh-privatekey
              path: id_rsa
            - key: ssh-publickey
              path: id_rsa.pub
            - key: ssh-publickey
              path: authorized_keys
            secretName: blake-ssh-secret
        - name: nnf
          hostPath:
            path: /mnt/nnf
            type: DirectoryOrCreate
        - name: kube-api-access-c5gl9
          projected:
            defaultMode: 420
            sources:
            - serviceAccountToken:
                expirationSeconds: 3607
                path: token
            - configMap:
                items:
                - key: ca.crt
                  path: ca.crt
                name: kube-root-ca.crt
            - downwardAPI:
                items:
                - fieldRef:
                    apiVersion: v1
                    fieldPath: metadata.namespace
                  path: namespace


---
apiVersion: batch/v1
kind: Job
metadata:
  name: blake-container3
  namespace: nnf-system
spec:
  # backoffLimit: 0
  # activeDeadlineSeconds: 0
  # ttlSecondsAfterFinished: 10
  # completions: 0
  # parallelism: 1
  template:
    metadata:
      labels:
        cray.nnf.container.workflow1: "true"
    spec:
      # restartPolicy: OnFailure
      restartPolicy: Never
      shareProcessNamespace: true
      hostname: blake-container3
      subdomain: blake-container
      containers:
        - name: container
          image: ghcr.io/nearnodeflash/nnf-mfu:82aa3c4b6655433dacbe3294ad359161c62219c6
          env:
            - name: OMPI_MCA_orte_keep_fqdn_hostnames
              value: "true"
            - name: MPIHOSTS
              value: ""
          command: 
            - /usr/sbin/sshd
          args:
            - -De
          securityContext:
            privileged: true
            capabilities:
              add: ["SETUID", "SETGID", "MKNOD"]
          volumeMounts:
            - name: ssh-auth
              mountPath: /root/.ssh
            # - name: kube-api-access-c5gl9
            #   mountPath: /var/run/secrets/kubernetes.io/serviceaccount
            #   readOnly: true
      tolerations:
        - effect: NoSchedule
          key: cray.nnf.node
          operator: Equal
          value: "true"
      nodeSelector:
        kubernetes.io/hostname: "kind-worker3"
      volumes:
        - name: ssh-auth
          secret:
            defaultMode: 384
            items:
            - key: ssh-privatekey
              path: id_rsa
            - key: ssh-publickey
              path: id_rsa.pub
            - key: ssh-publickey
              path: authorized_keys
            secretName: blake-ssh-secret
        - name: nnf
          hostPath:
            path: /mnt/nnf
            type: DirectoryOrCreate
        - name: kube-api-access-c5gl9
          projected:
            defaultMode: 420
            sources:
            - serviceAccountToken:
                expirationSeconds: 3607
                path: token
            - configMap:
                items:
                - key: ca.crt
                  path: ca.crt
                name: kube-root-ca.crt
            - downwardAPI:
                items:
                - fieldRef:
                    apiVersion: v1
                    fieldPath: metadata.namespace
                  path: namespace