---
apiVersion: v1
items:
- apiVersion: v1
  kind: Pod
  metadata:
    annotations:
      k8s.ovn.org/pod-networks: '{"default":{"ip_addresses":["10.133.0.35/23"],"mac_address":"0a:58:0a:85:00:23","gateway_ips":["10.133.0.1"],"routes":[{"dest":"10.132.0.0/14","nextHop":"10.133.0.1"},{"dest":"172.31.0.0/16","nextHop":"10.133.0.1"},{"dest":"169.254.0.5/32","nextHop":"10.133.0.1"},{"dest":"100.64.0.0/16","nextHop":"10.133.0.1"}],"ip_address":"10.133.0.35/23","gateway_ip":"10.133.0.1","role":"primary"}}'
      k8s.v1.cni.cncf.io/network-status: |-
        [{
            "name": "ovn-kubernetes",
            "interface": "eth0",
            "ips": [
                "10.133.0.35"
            ],
            "mac": "0a:58:0a:85:00:23",
            "default": true,
            "dns": {}
        }]
      openshift.io/scc: restricted-v2
      seccomp.security.alpha.kubernetes.io/pod: runtime/default
      security.openshift.io/validated-scc-subject-type: user
    creationTimestamp: "2026-11-02T12:06:49Z"
    generateName: e2e-distinct-2-simulated-kserve-cbd4f5586-
    generation: 1
    labels:
      app.kubernetes.io/component: llminferenceservice-workload
      app.kubernetes.io/name: e2e-distinct-2-simulated
      app.kubernetes.io/part-of: llminferenceservice
      kserve.io/component: workload
      llm-d.ai/role: both
      pod-template-hash: cbd4f5586
      topology.kubernetes.io/region: us-west-1
      topology.kubernetes.io/zone: us-west-1a
    managedFields:
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            f:k8s.ovn.org/pod-networks: {}
      manager: ip-10-0-133-239
      operation: Update
      subresource: status
      time: "2026-11-02T12:06:49Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:generateName: {}
          f:labels:
            .: {}
            f:app.kubernetes.io/component: {}
            f:app.kubernetes.io/name: {}
            f:app.kubernetes.io/part-of: {}
            f:kserve.io/component: {}
            f:llm-d.ai/role: {}
            f:pod-template-hash: {}
          f:ownerReferences:
            .: {}
            k:{"uid":"be0b99ab-dc8a-48fa-bdee-094c1a7f6652"}: {}
        f:spec:
          f:containers:
            k:{"name":"main"}:
              .: {}
              f:args: {}
              f:command: {}
              f:env:
                .: {}
                k:{"name":"HF_HUB_CACHE"}:
                  .: {}
                  f:name: {}
                  f:value: {}
                k:{"name":"HOME"}:
                  .: {}
                  f:name: {}
                  f:value: {}
                k:{"name":"POD_IP"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"POD_NAME"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"POD_NAMESPACE"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"VLLM_LOGGING_LEVEL"}:
                  .: {}
                  f:name: {}
                  f:value: {}
              f:image: {}
              f:imagePullPolicy: {}
              f:lifecycle:
                .: {}
                f:preStop:
                  .: {}
                  f:exec:
                    .: {}
                    f:command: {}
              f:livenessProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:name: {}
              f:ports:
                .: {}
                k:{"containerPort":8000,"protocol":"TCP"}:
                  .: {}
                  f:containerPort: {}
                  f:name: {}
                  f:protocol: {}
              f:readinessProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:resources:
                .: {}
                f:limits:
                  .: {}
                  f:cpu: {}
                  f:memory: {}
                f:requests:
                  .: {}
                  f:cpu: {}
                  f:memory: {}
              f:securityContext:
                .: {}
                f:allowPrivilegeEscalation: {}
                f:capabilities:
                  .: {}
                  f:drop: {}
                f:readOnlyRootFilesystem: {}
                f:runAsNonRoot: {}
                f:seccompProfile:
                  .: {}
                  f:type: {}
              f:startupProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:terminationMessagePath: {}
              f:terminationMessagePolicy: {}
              f:volumeMounts:
                .: {}
                k:{"mountPath":"/dev/shm"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/home"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/models"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/tmp"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/var/run/kserve/tls"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                  f:readOnly: {}
          f:dnsPolicy: {}
          f:enableServiceLinks: {}
          f:restartPolicy: {}
          f:schedulerName: {}
          f:securityContext: {}
          f:terminationGracePeriodSeconds: {}
          f:volumes:
            .: {}
            k:{"name":"dshm"}:
              .: {}
              f:emptyDir:
                .: {}
                f:medium: {}
                f:sizeLimit: {}
              f:name: {}
            k:{"name":"home"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
            k:{"name":"model-cache"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
            k:{"name":"tls-certs"}:
              .: {}
              f:name: {}
              f:secret:
                .: {}
                f:defaultMode: {}
                f:secretName: {}
            k:{"name":"tmp-dir"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
      manager: kube-controller-manager
      operation: Update
      time: "2026-11-02T12:06:49Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            f:k8s.v1.cni.cncf.io/network-status: {}
      manager: multus-daemon
      operation: Update
      subresource: status
      time: "2026-11-02T12:06:50Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:status:
          f:conditions:
            k:{"type":"ContainersReady"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"Initialized"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"PodReadyToStartContainers"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"PodScheduled"}:
              f:observedGeneration: {}
            k:{"type":"Ready"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
          f:containerStatuses: {}
          f:hostIP: {}
          f:hostIPs: {}
          f:observedGeneration: {}
          f:phase: {}
          f:podIP: {}
          f:podIPs:
            .: {}
            k:{"ip":"10.133.0.35"}:
              .: {}
              f:ip: {}
          f:startTime: {}
      manager: kubelet
      operation: Update
      subresource: status
      time: "2026-11-02T12:07:00Z"
    name: e2e-distinct-2-simulated-kserve-cbd4f5586-4l7p7
    namespace: llm
    ownerReferences:
    - apiVersion: apps/v1
      blockOwnerDeletion: true
      controller: true
      kind: ReplicaSet
      name: e2e-distinct-2-simulated-kserve-cbd4f5586
      uid: be0b99ab-dc8a-48fa-bdee-094c1a7f6652
    resourceVersion: "20274"
    uid: 6a1bdd91-9d36-4c44-a3ac-694afddf8ba0
  spec:
    containers:
    - args:
      - --port
      - "8000"
      - --model
      - test/e2e-distinct-model-2
      - --mode
      - random
      - --no-mm-encoder-only
      - --ssl-certfile
      - /var/run/kserve/tls/tls.crt
      - --ssl-keyfile
      - /var/run/kserve/tls/tls.key
      command:
      - /app/llm-d-inference-sim
      env:
      - name: POD_NAME
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: metadata.name
      - name: POD_NAMESPACE
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: metadata.namespace
      - name: POD_IP
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: status.podIP
      - name: HOME
        value: /home
      - name: VLLM_LOGGING_LEVEL
        value: INFO
      - name: HF_HUB_CACHE
        value: /models
      image: ghcr.io/llm-d/llm-d-inference-sim:v0.8.2
      imagePullPolicy: Always
      lifecycle:
        preStop:
          exec:
            command:
            - /bin/sleep
            - "15"
      livenessProbe:
        failureThreshold: 10
        httpGet:
          path: /health
          port: https
          scheme: HTTPS
        periodSeconds: 10
        successThreshold: 1
        timeoutSeconds: 1
      name: main
      ports:
      - containerPort: 8000
        name: https
        protocol: TCP
      readinessProbe:
        failureThreshold: 2
        httpGet:
          path: /ready
          port: https
          scheme: HTTPS
        periodSeconds: 1
        successThreshold: 1
        timeoutSeconds: 1
      resources:
        limits:
          cpu: 500m
          memory: 512Mi
        requests:
          cpu: 100m
          memory: 256Mi
      securityContext:
        allowPrivilegeEscalation: false
        capabilities:
          drop:
          - ALL
        readOnlyRootFilesystem: false
        runAsNonRoot: true
        runAsUser: 1000710000
        seccompProfile:
          type: RuntimeDefault
      startupProbe:
        failureThreshold: 60
        httpGet:
          path: /health
          port: 8000
          scheme: HTTPS
        periodSeconds: 10
        successThreshold: 1
        timeoutSeconds: 1
      terminationMessagePath: /dev/termination-log
      terminationMessagePolicy: FallbackToLogsOnError
      volumeMounts:
      - mountPath: /home
        name: home
      - mountPath: /tmp
        name: tmp-dir
      - mountPath: /dev/shm
        name: dshm
      - mountPath: /models
        name: model-cache
      - mountPath: /var/run/kserve/tls
        name: tls-certs
        readOnly: true
      - mountPath: /var/run/secrets/kubernetes.io/serviceaccount
        name: kube-api-access-6mtmh
        readOnly: true
    dnsPolicy: ClusterFirst
    enableServiceLinks: true
    imagePullSecrets:
    - name: default-dockercfg-p8m6b
    nodeName: ip-10-0-133-239.us-west-1.compute.internal
    preemptionPolicy: PreemptLowerPriority
    priority: 0
    restartPolicy: Always
    schedulerName: default-scheduler
    securityContext:
      fsGroup: 1000710000
      seLinuxOptions:
        level: s0:c27,c4
      seccompProfile:
        type: RuntimeDefault
    serviceAccount: default
    serviceAccountName: default
    terminationGracePeriodSeconds: 60
    tolerations:
    - effect: NoExecute
      key: node.kubernetes.io/not-ready
      operator: Exists
      tolerationSeconds: 300
    - effect: NoExecute
      key: node.kubernetes.io/unreachable
      operator: Exists
      tolerationSeconds: 300
    - effect: NoSchedule
      key: node.kubernetes.io/memory-pressure
      operator: Exists
    volumes:
    - emptyDir: {}
      name: home
    - emptyDir:
        medium: Memory
        sizeLimit: 1Gi
      name: dshm
    - emptyDir: {}
      name: model-cache
    - emptyDir: {}
      name: tmp-dir
    - name: tls-certs
      secret:
        defaultMode: 420
        secretName: e2e-distinct-2-simulated-kserve-self-signed-certs
    - name: kube-api-access-6mtmh
      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
        - configMap:
            items:
            - key: service-ca.crt
              path: service-ca.crt
            name: openshift-service-ca.crt
  status:
    conditions:
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:06:53Z"
      observedGeneration: 1
      status: "True"
      type: PodReadyToStartContainers
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:06:49Z"
      observedGeneration: 1
      status: "True"
      type: Initialized
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:00Z"
      observedGeneration: 1
      status: "True"
      type: Ready
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:00Z"
      observedGeneration: 1
      status: "True"
      type: ContainersReady
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:06:49Z"
      observedGeneration: 1
      status: "True"
      type: PodScheduled
    containerStatuses:
    - allocatedResources:
        cpu: 100m
        memory: 256Mi
      containerID: cri-o://163097f05ebe4550775ce062df9bdcb19facace79b9d54e9985f559717b26484
      image: ghcr.io/llm-d/llm-d-inference-sim:v0.8.2
      imageID: ghcr.io/llm-d/llm-d-inference-sim@sha256:bab162bd25e2ed8b15022387cdb223023aeb33be49476af9f0115c0398fb8ff5
      lastState: {}
      name: main
      ready: true
      resources:
        limits:
          cpu: 500m
          memory: 512Mi
        requests:
          cpu: 100m
          memory: 256Mi
      restartCount: 0
      started: true
      state:
        running:
          startedAt: "2026-11-02T12:06:52Z"
      user:
        linux:
          gid: 0
          supplementalGroups:
          - 0
          - 1000710000
          uid: 1000710000
      volumeMounts:
      - mountPath: /home
        name: home
      - mountPath: /tmp
        name: tmp-dir
      - mountPath: /dev/shm
        name: dshm
      - mountPath: /models
        name: model-cache
      - mountPath: /var/run/kserve/tls
        name: tls-certs
        readOnly: true
        recursiveReadOnly: Disabled
      - mountPath: /var/run/secrets/kubernetes.io/serviceaccount
        name: kube-api-access-6mtmh
        readOnly: true
        recursiveReadOnly: Disabled
    hostIP: 10.0.133.239
    hostIPs:
    - ip: 10.0.133.239
    observedGeneration: 1
    phase: Running
    podIP: 10.133.0.35
    podIPs:
    - ip: 10.133.0.35
    qosClass: Burstable
    startTime: "2026-11-02T12:06:49Z"
- apiVersion: v1
  kind: Pod
  metadata:
    annotations:
      k8s.ovn.org/pod-networks: '{"default":{"ip_addresses":["10.133.0.36/23"],"mac_address":"0a:58:0a:85:00:24","gateway_ips":["10.133.0.1"],"routes":[{"dest":"10.132.0.0/14","nextHop":"10.133.0.1"},{"dest":"172.31.0.0/16","nextHop":"10.133.0.1"},{"dest":"169.254.0.5/32","nextHop":"10.133.0.1"},{"dest":"100.64.0.0/16","nextHop":"10.133.0.1"}],"ip_address":"10.133.0.36/23","gateway_ip":"10.133.0.1","role":"primary"}}'
      k8s.v1.cni.cncf.io/network-status: |-
        [{
            "name": "ovn-kubernetes",
            "interface": "eth0",
            "ips": [
                "10.133.0.36"
            ],
            "mac": "0a:58:0a:85:00:24",
            "default": true,
            "dns": {}
        }]
      openshift.io/scc: restricted-v2
      seccomp.security.alpha.kubernetes.io/pod: runtime/default
      security.openshift.io/validated-scc-subject-type: user
    creationTimestamp: "2026-11-02T12:06:55Z"
    generateName: e2e-distinct-simulated-kserve-7b998b79d9-
    generation: 1
    labels:
      app.kubernetes.io/component: llminferenceservice-workload
      app.kubernetes.io/name: e2e-distinct-simulated
      app.kubernetes.io/part-of: llminferenceservice
      kserve.io/component: workload
      llm-d.ai/role: both
      pod-template-hash: 7b998b79d9
      topology.kubernetes.io/region: us-west-1
      topology.kubernetes.io/zone: us-west-1a
    managedFields:
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            f:k8s.ovn.org/pod-networks: {}
      manager: ip-10-0-133-239
      operation: Update
      subresource: status
      time: "2026-11-02T12:06:55Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:generateName: {}
          f:labels:
            .: {}
            f:app.kubernetes.io/component: {}
            f:app.kubernetes.io/name: {}
            f:app.kubernetes.io/part-of: {}
            f:kserve.io/component: {}
            f:llm-d.ai/role: {}
            f:pod-template-hash: {}
          f:ownerReferences:
            .: {}
            k:{"uid":"399292fa-d2f3-476d-a3ff-c3d212488a70"}: {}
        f:spec:
          f:containers:
            k:{"name":"main"}:
              .: {}
              f:args: {}
              f:command: {}
              f:env:
                .: {}
                k:{"name":"HF_HUB_CACHE"}:
                  .: {}
                  f:name: {}
                  f:value: {}
                k:{"name":"HOME"}:
                  .: {}
                  f:name: {}
                  f:value: {}
                k:{"name":"POD_IP"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"POD_NAME"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"POD_NAMESPACE"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"VLLM_LOGGING_LEVEL"}:
                  .: {}
                  f:name: {}
                  f:value: {}
              f:image: {}
              f:imagePullPolicy: {}
              f:lifecycle:
                .: {}
                f:preStop:
                  .: {}
                  f:exec:
                    .: {}
                    f:command: {}
              f:livenessProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:name: {}
              f:ports:
                .: {}
                k:{"containerPort":8000,"protocol":"TCP"}:
                  .: {}
                  f:containerPort: {}
                  f:name: {}
                  f:protocol: {}
              f:readinessProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:resources:
                .: {}
                f:limits:
                  .: {}
                  f:cpu: {}
                  f:memory: {}
                f:requests:
                  .: {}
                  f:cpu: {}
                  f:memory: {}
              f:securityContext:
                .: {}
                f:allowPrivilegeEscalation: {}
                f:capabilities:
                  .: {}
                  f:drop: {}
                f:readOnlyRootFilesystem: {}
                f:runAsNonRoot: {}
                f:seccompProfile:
                  .: {}
                  f:type: {}
              f:startupProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:terminationMessagePath: {}
              f:terminationMessagePolicy: {}
              f:volumeMounts:
                .: {}
                k:{"mountPath":"/dev/shm"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/home"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/models"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/tmp"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/var/run/kserve/tls"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                  f:readOnly: {}
          f:dnsPolicy: {}
          f:enableServiceLinks: {}
          f:restartPolicy: {}
          f:schedulerName: {}
          f:securityContext: {}
          f:terminationGracePeriodSeconds: {}
          f:volumes:
            .: {}
            k:{"name":"dshm"}:
              .: {}
              f:emptyDir:
                .: {}
                f:medium: {}
                f:sizeLimit: {}
              f:name: {}
            k:{"name":"home"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
            k:{"name":"model-cache"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
            k:{"name":"tls-certs"}:
              .: {}
              f:name: {}
              f:secret:
                .: {}
                f:defaultMode: {}
                f:secretName: {}
            k:{"name":"tmp-dir"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
      manager: kube-controller-manager
      operation: Update
      time: "2026-11-02T12:06:55Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            f:k8s.v1.cni.cncf.io/network-status: {}
      manager: multus-daemon
      operation: Update
      subresource: status
      time: "2026-11-02T12:06:56Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:status:
          f:conditions:
            k:{"type":"ContainersReady"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"Initialized"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"PodReadyToStartContainers"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"PodScheduled"}:
              f:observedGeneration: {}
            k:{"type":"Ready"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
          f:containerStatuses: {}
          f:hostIP: {}
          f:hostIPs: {}
          f:observedGeneration: {}
          f:phase: {}
          f:podIP: {}
          f:podIPs:
            .: {}
            k:{"ip":"10.133.0.36"}:
              .: {}
              f:ip: {}
          f:startTime: {}
      manager: kubelet
      operation: Update
      subresource: status
      time: "2026-11-02T12:07:06Z"
    name: e2e-distinct-simulated-kserve-7b998b79d9-8j57g
    namespace: llm
    ownerReferences:
    - apiVersion: apps/v1
      blockOwnerDeletion: true
      controller: true
      kind: ReplicaSet
      name: e2e-distinct-simulated-kserve-7b998b79d9
      uid: 399292fa-d2f3-476d-a3ff-c3d212488a70
    resourceVersion: "20423"
    uid: 53517c94-5480-4ee7-abcf-cdc03f8af0a1
  spec:
    containers:
    - args:
      - --port
      - "8000"
      - --model
      - test/e2e-distinct-model
      - --mode
      - random
      - --no-mm-encoder-only
      - --ssl-certfile
      - /var/run/kserve/tls/tls.crt
      - --ssl-keyfile
      - /var/run/kserve/tls/tls.key
      command:
      - /app/llm-d-inference-sim
      env:
      - name: POD_NAME
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: metadata.name
      - name: POD_NAMESPACE
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: metadata.namespace
      - name: POD_IP
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: status.podIP
      - name: HOME
        value: /home
      - name: VLLM_LOGGING_LEVEL
        value: INFO
      - name: HF_HUB_CACHE
        value: /models
      image: ghcr.io/llm-d/llm-d-inference-sim:v0.8.2
      imagePullPolicy: Always
      lifecycle:
        preStop:
          exec:
            command:
            - /bin/sleep
            - "15"
      livenessProbe:
        failureThreshold: 10
        httpGet:
          path: /health
          port: https
          scheme: HTTPS
        periodSeconds: 10
        successThreshold: 1
        timeoutSeconds: 1
      name: main
      ports:
      - containerPort: 8000
        name: https
        protocol: TCP
      readinessProbe:
        failureThreshold: 2
        httpGet:
          path: /ready
          port: https
          scheme: HTTPS
        periodSeconds: 1
        successThreshold: 1
        timeoutSeconds: 1
      resources:
        limits:
          cpu: 500m
          memory: 512Mi
        requests:
          cpu: 100m
          memory: 256Mi
      securityContext:
        allowPrivilegeEscalation: false
        capabilities:
          drop:
          - ALL
        readOnlyRootFilesystem: false
        runAsNonRoot: true
        runAsUser: 1000710000
        seccompProfile:
          type: RuntimeDefault
      startupProbe:
        failureThreshold: 60
        httpGet:
          path: /health
          port: 8000
          scheme: HTTPS
        periodSeconds: 10
        successThreshold: 1
        timeoutSeconds: 1
      terminationMessagePath: /dev/termination-log
      terminationMessagePolicy: FallbackToLogsOnError
      volumeMounts:
      - mountPath: /home
        name: home
      - mountPath: /tmp
        name: tmp-dir
      - mountPath: /dev/shm
        name: dshm
      - mountPath: /models
        name: model-cache
      - mountPath: /var/run/kserve/tls
        name: tls-certs
        readOnly: true
      - mountPath: /var/run/secrets/kubernetes.io/serviceaccount
        name: kube-api-access-tzn9t
        readOnly: true
    dnsPolicy: ClusterFirst
    enableServiceLinks: true
    imagePullSecrets:
    - name: default-dockercfg-p8m6b
    nodeName: ip-10-0-133-239.us-west-1.compute.internal
    preemptionPolicy: PreemptLowerPriority
    priority: 0
    restartPolicy: Always
    schedulerName: default-scheduler
    securityContext:
      fsGroup: 1000710000
      seLinuxOptions:
        level: s0:c27,c4
      seccompProfile:
        type: RuntimeDefault
    serviceAccount: default
    serviceAccountName: default
    terminationGracePeriodSeconds: 60
    tolerations:
    - effect: NoExecute
      key: node.kubernetes.io/not-ready
      operator: Exists
      tolerationSeconds: 300
    - effect: NoExecute
      key: node.kubernetes.io/unreachable
      operator: Exists
      tolerationSeconds: 300
    - effect: NoSchedule
      key: node.kubernetes.io/memory-pressure
      operator: Exists
    volumes:
    - emptyDir: {}
      name: home
    - emptyDir:
        medium: Memory
        sizeLimit: 1Gi
      name: dshm
    - emptyDir: {}
      name: model-cache
    - emptyDir: {}
      name: tmp-dir
    - name: tls-certs
      secret:
        defaultMode: 420
        secretName: e2e-distinct-simulated-kserve-self-signed-certs
    - name: kube-api-access-tzn9t
      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
        - configMap:
            items:
            - key: service-ca.crt
              path: service-ca.crt
            name: openshift-service-ca.crt
  status:
    conditions:
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:06:57Z"
      observedGeneration: 1
      status: "True"
      type: PodReadyToStartContainers
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:06:55Z"
      observedGeneration: 1
      status: "True"
      type: Initialized
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:06Z"
      observedGeneration: 1
      status: "True"
      type: Ready
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:06Z"
      observedGeneration: 1
      status: "True"
      type: ContainersReady
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:06:55Z"
      observedGeneration: 1
      status: "True"
      type: PodScheduled
    containerStatuses:
    - allocatedResources:
        cpu: 100m
        memory: 256Mi
      containerID: cri-o://d3b661b64e838f737339b9db9f30f01819afdcdf1e56017b5464c610a061847c
      image: ghcr.io/llm-d/llm-d-inference-sim:v0.8.2
      imageID: ghcr.io/llm-d/llm-d-inference-sim@sha256:bab162bd25e2ed8b15022387cdb223023aeb33be49476af9f0115c0398fb8ff5
      lastState: {}
      name: main
      ready: true
      resources:
        limits:
          cpu: 500m
          memory: 512Mi
        requests:
          cpu: 100m
          memory: 256Mi
      restartCount: 0
      started: true
      state:
        running:
          startedAt: "2026-11-02T12:06:56Z"
      user:
        linux:
          gid: 0
          supplementalGroups:
          - 0
          - 1000710000
          uid: 1000710000
      volumeMounts:
      - mountPath: /home
        name: home
      - mountPath: /tmp
        name: tmp-dir
      - mountPath: /dev/shm
        name: dshm
      - mountPath: /models
        name: model-cache
      - mountPath: /var/run/kserve/tls
        name: tls-certs
        readOnly: true
        recursiveReadOnly: Disabled
      - mountPath: /var/run/secrets/kubernetes.io/serviceaccount
        name: kube-api-access-tzn9t
        readOnly: true
        recursiveReadOnly: Disabled
    hostIP: 10.0.133.239
    hostIPs:
    - ip: 10.0.133.239
    observedGeneration: 1
    phase: Running
    podIP: 10.133.0.36
    podIPs:
    - ip: 10.133.0.36
    qosClass: Burstable
    startTime: "2026-11-02T12:06:55Z"
- apiVersion: v1
  kind: Pod
  metadata:
    annotations:
      k8s.ovn.org/pod-networks: '{"default":{"ip_addresses":["10.133.0.37/23"],"mac_address":"0a:58:0a:85:00:25","gateway_ips":["10.133.0.1"],"routes":[{"dest":"10.132.0.0/14","nextHop":"10.133.0.1"},{"dest":"172.31.0.0/16","nextHop":"10.133.0.1"},{"dest":"169.254.0.5/32","nextHop":"10.133.0.1"},{"dest":"100.64.0.0/16","nextHop":"10.133.0.1"}],"ip_address":"10.133.0.37/23","gateway_ip":"10.133.0.1","role":"primary"}}'
      k8s.v1.cni.cncf.io/network-status: |-
        [{
            "name": "ovn-kubernetes",
            "interface": "eth0",
            "ips": [
                "10.133.0.37"
            ],
            "mac": "0a:58:0a:85:00:25",
            "default": true,
            "dns": {}
        }]
      openshift.io/scc: restricted-v2
      seccomp.security.alpha.kubernetes.io/pod: runtime/default
      security.openshift.io/validated-scc-subject-type: user
    creationTimestamp: "2026-11-02T12:07:00Z"
    generateName: e2e-embedding-simulated-kserve-67bf594cd6-
    generation: 1
    labels:
      app.kubernetes.io/component: llminferenceservice-workload
      app.kubernetes.io/name: e2e-embedding-simulated
      app.kubernetes.io/part-of: llminferenceservice
      kserve.io/component: workload
      llm-d.ai/role: both
      pod-template-hash: 67bf594cd6
      topology.kubernetes.io/region: us-west-1
      topology.kubernetes.io/zone: us-west-1a
    managedFields:
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            f:k8s.ovn.org/pod-networks: {}
      manager: ip-10-0-133-239
      operation: Update
      subresource: status
      time: "2026-11-02T12:07:00Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:generateName: {}
          f:labels:
            .: {}
            f:app.kubernetes.io/component: {}
            f:app.kubernetes.io/name: {}
            f:app.kubernetes.io/part-of: {}
            f:kserve.io/component: {}
            f:llm-d.ai/role: {}
            f:pod-template-hash: {}
          f:ownerReferences:
            .: {}
            k:{"uid":"20bbfe6f-c777-4a4f-97e8-5bcb26027628"}: {}
        f:spec:
          f:containers:
            k:{"name":"main"}:
              .: {}
              f:args: {}
              f:command: {}
              f:env:
                .: {}
                k:{"name":"HF_HUB_CACHE"}:
                  .: {}
                  f:name: {}
                  f:value: {}
                k:{"name":"HOME"}:
                  .: {}
                  f:name: {}
                  f:value: {}
                k:{"name":"POD_IP"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"POD_NAME"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"POD_NAMESPACE"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"VLLM_LOGGING_LEVEL"}:
                  .: {}
                  f:name: {}
                  f:value: {}
              f:image: {}
              f:imagePullPolicy: {}
              f:lifecycle:
                .: {}
                f:preStop:
                  .: {}
                  f:exec:
                    .: {}
                    f:command: {}
              f:livenessProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:name: {}
              f:ports:
                .: {}
                k:{"containerPort":8000,"protocol":"TCP"}:
                  .: {}
                  f:containerPort: {}
                  f:name: {}
                  f:protocol: {}
              f:readinessProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:resources:
                .: {}
                f:limits:
                  .: {}
                  f:cpu: {}
                  f:memory: {}
                f:requests:
                  .: {}
                  f:cpu: {}
                  f:memory: {}
              f:securityContext:
                .: {}
                f:allowPrivilegeEscalation: {}
                f:capabilities:
                  .: {}
                  f:drop: {}
                f:readOnlyRootFilesystem: {}
                f:runAsNonRoot: {}
                f:seccompProfile:
                  .: {}
                  f:type: {}
              f:startupProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:terminationMessagePath: {}
              f:terminationMessagePolicy: {}
              f:volumeMounts:
                .: {}
                k:{"mountPath":"/dev/shm"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/home"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/models"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/tmp"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/var/run/kserve/tls"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                  f:readOnly: {}
          f:dnsPolicy: {}
          f:enableServiceLinks: {}
          f:restartPolicy: {}
          f:schedulerName: {}
          f:securityContext: {}
          f:terminationGracePeriodSeconds: {}
          f:volumes:
            .: {}
            k:{"name":"dshm"}:
              .: {}
              f:emptyDir:
                .: {}
                f:medium: {}
                f:sizeLimit: {}
              f:name: {}
            k:{"name":"home"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
            k:{"name":"model-cache"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
            k:{"name":"tls-certs"}:
              .: {}
              f:name: {}
              f:secret:
                .: {}
                f:defaultMode: {}
                f:secretName: {}
            k:{"name":"tmp-dir"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
      manager: kube-controller-manager
      operation: Update
      time: "2026-11-02T12:07:00Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            f:k8s.v1.cni.cncf.io/network-status: {}
      manager: multus-daemon
      operation: Update
      subresource: status
      time: "2026-11-02T12:07:00Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:status:
          f:conditions:
            k:{"type":"ContainersReady"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"Initialized"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"PodReadyToStartContainers"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"PodScheduled"}:
              f:observedGeneration: {}
            k:{"type":"Ready"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
          f:containerStatuses: {}
          f:hostIP: {}
          f:hostIPs: {}
          f:observedGeneration: {}
          f:phase: {}
          f:podIP: {}
          f:podIPs:
            .: {}
            k:{"ip":"10.133.0.37"}:
              .: {}
              f:ip: {}
          f:startTime: {}
      manager: kubelet
      operation: Update
      subresource: status
      time: "2026-11-02T12:07:11Z"
    name: e2e-embedding-simulated-kserve-67bf594cd6-r8qcw
    namespace: llm
    ownerReferences:
    - apiVersion: apps/v1
      blockOwnerDeletion: true
      controller: true
      kind: ReplicaSet
      name: e2e-embedding-simulated-kserve-67bf594cd6
      uid: 20bbfe6f-c777-4a4f-97e8-5bcb26027628
    resourceVersion: "20558"
    uid: 55c46ddc-9659-48d9-92c6-d8363f573813
  spec:
    containers:
    - args:
      - --port
      - "8000"
      - --model
      - test/e2e-embedding-model
      - --mode
      - random
      - --no-mm-encoder-only
      - --ssl-certfile
      - /var/run/kserve/tls/tls.crt
      - --ssl-keyfile
      - /var/run/kserve/tls/tls.key
      command:
      - /app/llm-d-inference-sim
      env:
      - name: POD_NAME
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: metadata.name
      - name: POD_NAMESPACE
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: metadata.namespace
      - name: POD_IP
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: status.podIP
      - name: HOME
        value: /home
      - name: VLLM_LOGGING_LEVEL
        value: INFO
      - name: HF_HUB_CACHE
        value: /models
      image: ghcr.io/llm-d/llm-d-inference-sim:v0.8.2
      imagePullPolicy: Always
      lifecycle:
        preStop:
          exec:
            command:
            - /bin/sleep
            - "15"
      livenessProbe:
        failureThreshold: 10
        httpGet:
          path: /health
          port: https
          scheme: HTTPS
        periodSeconds: 10
        successThreshold: 1
        timeoutSeconds: 1
      name: main
      ports:
      - containerPort: 8000
        name: https
        protocol: TCP
      readinessProbe:
        failureThreshold: 2
        httpGet:
          path: /ready
          port: https
          scheme: HTTPS
        periodSeconds: 1
        successThreshold: 1
        timeoutSeconds: 1
      resources:
        limits:
          cpu: 500m
          memory: 512Mi
        requests:
          cpu: 100m
          memory: 256Mi
      securityContext:
        allowPrivilegeEscalation: false
        capabilities:
          drop:
          - ALL
        readOnlyRootFilesystem: false
        runAsNonRoot: true
        runAsUser: 1000710000
        seccompProfile:
          type: RuntimeDefault
      startupProbe:
        failureThreshold: 60
        httpGet:
          path: /health
          port: 8000
          scheme: HTTPS
        periodSeconds: 10
        successThreshold: 1
        timeoutSeconds: 1
      terminationMessagePath: /dev/termination-log
      terminationMessagePolicy: FallbackToLogsOnError
      volumeMounts:
      - mountPath: /home
        name: home
      - mountPath: /tmp
        name: tmp-dir
      - mountPath: /dev/shm
        name: dshm
      - mountPath: /models
        name: model-cache
      - mountPath: /var/run/kserve/tls
        name: tls-certs
        readOnly: true
      - mountPath: /var/run/secrets/kubernetes.io/serviceaccount
        name: kube-api-access-8wd2k
        readOnly: true
    dnsPolicy: ClusterFirst
    enableServiceLinks: true
    imagePullSecrets:
    - name: default-dockercfg-p8m6b
    nodeName: ip-10-0-133-239.us-west-1.compute.internal
    preemptionPolicy: PreemptLowerPriority
    priority: 0
    restartPolicy: Always
    schedulerName: default-scheduler
    securityContext:
      fsGroup: 1000710000
      seLinuxOptions:
        level: s0:c27,c4
      seccompProfile:
        type: RuntimeDefault
    serviceAccount: default
    serviceAccountName: default
    terminationGracePeriodSeconds: 60
    tolerations:
    - effect: NoExecute
      key: node.kubernetes.io/not-ready
      operator: Exists
      tolerationSeconds: 300
    - effect: NoExecute
      key: node.kubernetes.io/unreachable
      operator: Exists
      tolerationSeconds: 300
    - effect: NoSchedule
      key: node.kubernetes.io/memory-pressure
      operator: Exists
    volumes:
    - emptyDir: {}
      name: home
    - emptyDir:
        medium: Memory
        sizeLimit: 1Gi
      name: dshm
    - emptyDir: {}
      name: model-cache
    - emptyDir: {}
      name: tmp-dir
    - name: tls-certs
      secret:
        defaultMode: 420
        secretName: e2e-embedding-simulated-kserve-self-signed-certs
    - name: kube-api-access-8wd2k
      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
        - configMap:
            items:
            - key: service-ca.crt
              path: service-ca.crt
            name: openshift-service-ca.crt
  status:
    conditions:
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:02Z"
      observedGeneration: 1
      status: "True"
      type: PodReadyToStartContainers
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:00Z"
      observedGeneration: 1
      status: "True"
      type: Initialized
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:11Z"
      observedGeneration: 1
      status: "True"
      type: Ready
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:11Z"
      observedGeneration: 1
      status: "True"
      type: ContainersReady
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:00Z"
      observedGeneration: 1
      status: "True"
      type: PodScheduled
    containerStatuses:
    - allocatedResources:
        cpu: 100m
        memory: 256Mi
      containerID: cri-o://3d34f7edb7c9c52838e37bdd5e547035e10d7aa19920acd8a920331e2b1a284c
      image: ghcr.io/llm-d/llm-d-inference-sim:v0.8.2
      imageID: ghcr.io/llm-d/llm-d-inference-sim@sha256:bab162bd25e2ed8b15022387cdb223023aeb33be49476af9f0115c0398fb8ff5
      lastState: {}
      name: main
      ready: true
      resources:
        limits:
          cpu: 500m
          memory: 512Mi
        requests:
          cpu: 100m
          memory: 256Mi
      restartCount: 0
      started: true
      state:
        running:
          startedAt: "2026-11-02T12:07:01Z"
      user:
        linux:
          gid: 0
          supplementalGroups:
          - 0
          - 1000710000
          uid: 1000710000
      volumeMounts:
      - mountPath: /home
        name: home
      - mountPath: /tmp
        name: tmp-dir
      - mountPath: /dev/shm
        name: dshm
      - mountPath: /models
        name: model-cache
      - mountPath: /var/run/kserve/tls
        name: tls-certs
        readOnly: true
        recursiveReadOnly: Disabled
      - mountPath: /var/run/secrets/kubernetes.io/serviceaccount
        name: kube-api-access-8wd2k
        readOnly: true
        recursiveReadOnly: Disabled
    hostIP: 10.0.133.239
    hostIPs:
    - ip: 10.0.133.239
    observedGeneration: 1
    phase: Running
    podIP: 10.133.0.37
    podIPs:
    - ip: 10.133.0.37
    qosClass: Burstable
    startTime: "2026-11-02T12:07:00Z"
- apiVersion: v1
  kind: Pod
  metadata:
    annotations:
      k8s.ovn.org/pod-networks: '{"default":{"ip_addresses":["10.132.0.49/23"],"mac_address":"0a:58:0a:84:00:31","gateway_ips":["10.132.0.1"],"routes":[{"dest":"10.132.0.0/14","nextHop":"10.132.0.1"},{"dest":"172.31.0.0/16","nextHop":"10.132.0.1"},{"dest":"169.254.0.5/32","nextHop":"10.132.0.1"},{"dest":"100.64.0.0/16","nextHop":"10.132.0.1"}],"ip_address":"10.132.0.49/23","gateway_ip":"10.132.0.1","role":"primary"}}'
      k8s.v1.cni.cncf.io/network-status: |-
        [{
            "name": "ovn-kubernetes",
            "interface": "eth0",
            "ips": [
                "10.132.0.49"
            ],
            "mac": "0a:58:0a:84:00:31",
            "default": true,
            "dns": {}
        }]
      openshift.io/scc: restricted-v2
      seccomp.security.alpha.kubernetes.io/pod: runtime/default
      security.openshift.io/validated-scc-subject-type: user
    creationTimestamp: "2026-11-02T12:07:04Z"
    generateName: e2e-trlp-test-simulated-kserve-69b588897d-
    generation: 1
    labels:
      app.kubernetes.io/component: llminferenceservice-workload
      app.kubernetes.io/name: e2e-trlp-test-simulated
      app.kubernetes.io/part-of: llminferenceservice
      kserve.io/component: workload
      llm-d.ai/role: both
      pod-template-hash: 69b588897d
      topology.kubernetes.io/region: us-west-1
      topology.kubernetes.io/zone: us-west-1a
    managedFields:
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            f:k8s.ovn.org/pod-networks: {}
      manager: ip-10-0-135-24
      operation: Update
      subresource: status
      time: "2026-11-02T12:07:04Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:generateName: {}
          f:labels:
            .: {}
            f:app.kubernetes.io/component: {}
            f:app.kubernetes.io/name: {}
            f:app.kubernetes.io/part-of: {}
            f:kserve.io/component: {}
            f:llm-d.ai/role: {}
            f:pod-template-hash: {}
          f:ownerReferences:
            .: {}
            k:{"uid":"ea464174-4d4f-4bab-8197-eb44669a3f6a"}: {}
        f:spec:
          f:containers:
            k:{"name":"main"}:
              .: {}
              f:args: {}
              f:command: {}
              f:env:
                .: {}
                k:{"name":"HF_HUB_CACHE"}:
                  .: {}
                  f:name: {}
                  f:value: {}
                k:{"name":"HOME"}:
                  .: {}
                  f:name: {}
                  f:value: {}
                k:{"name":"POD_IP"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"POD_NAME"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"POD_NAMESPACE"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"VLLM_LOGGING_LEVEL"}:
                  .: {}
                  f:name: {}
                  f:value: {}
              f:image: {}
              f:imagePullPolicy: {}
              f:lifecycle:
                .: {}
                f:preStop:
                  .: {}
                  f:exec:
                    .: {}
                    f:command: {}
              f:livenessProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:name: {}
              f:ports:
                .: {}
                k:{"containerPort":8000,"protocol":"TCP"}:
                  .: {}
                  f:containerPort: {}
                  f:name: {}
                  f:protocol: {}
              f:readinessProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:resources:
                .: {}
                f:limits:
                  .: {}
                  f:cpu: {}
                  f:memory: {}
                f:requests:
                  .: {}
                  f:cpu: {}
                  f:memory: {}
              f:securityContext:
                .: {}
                f:allowPrivilegeEscalation: {}
                f:capabilities:
                  .: {}
                  f:drop: {}
                f:readOnlyRootFilesystem: {}
                f:runAsNonRoot: {}
                f:seccompProfile:
                  .: {}
                  f:type: {}
              f:startupProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:terminationMessagePath: {}
              f:terminationMessagePolicy: {}
              f:volumeMounts:
                .: {}
                k:{"mountPath":"/dev/shm"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/home"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/models"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/tmp"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/var/run/kserve/tls"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                  f:readOnly: {}
          f:dnsPolicy: {}
          f:enableServiceLinks: {}
          f:restartPolicy: {}
          f:schedulerName: {}
          f:securityContext: {}
          f:terminationGracePeriodSeconds: {}
          f:volumes:
            .: {}
            k:{"name":"dshm"}:
              .: {}
              f:emptyDir:
                .: {}
                f:medium: {}
                f:sizeLimit: {}
              f:name: {}
            k:{"name":"home"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
            k:{"name":"model-cache"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
            k:{"name":"tls-certs"}:
              .: {}
              f:name: {}
              f:secret:
                .: {}
                f:defaultMode: {}
                f:secretName: {}
            k:{"name":"tmp-dir"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
      manager: kube-controller-manager
      operation: Update
      time: "2026-11-02T12:07:04Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            f:k8s.v1.cni.cncf.io/network-status: {}
      manager: multus-daemon
      operation: Update
      subresource: status
      time: "2026-11-02T12:07:04Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:status:
          f:conditions:
            k:{"type":"ContainersReady"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"Initialized"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"PodReadyToStartContainers"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"PodScheduled"}:
              f:observedGeneration: {}
            k:{"type":"Ready"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
          f:containerStatuses: {}
          f:hostIP: {}
          f:hostIPs: {}
          f:observedGeneration: {}
          f:phase: {}
          f:podIP: {}
          f:podIPs:
            .: {}
            k:{"ip":"10.132.0.49"}:
              .: {}
              f:ip: {}
          f:startTime: {}
      manager: kubelet
      operation: Update
      subresource: status
      time: "2026-11-02T12:07:14Z"
    name: e2e-trlp-test-simulated-kserve-69b588897d-2hcp9
    namespace: llm
    ownerReferences:
    - apiVersion: apps/v1
      blockOwnerDeletion: true
      controller: true
      kind: ReplicaSet
      name: e2e-trlp-test-simulated-kserve-69b588897d
      uid: ea464174-4d4f-4bab-8197-eb44669a3f6a
    resourceVersion: "20691"
    uid: ded7be50-33c3-48ec-9c1e-cab1aaf939b1
  spec:
    containers:
    - args:
      - --port
      - "8000"
      - --model
      - test/e2e-trlp-test-model
      - --mode
      - random
      - --no-mm-encoder-only
      - --ssl-certfile
      - /var/run/kserve/tls/tls.crt
      - --ssl-keyfile
      - /var/run/kserve/tls/tls.key
      command:
      - /app/llm-d-inference-sim
      env:
      - name: POD_NAME
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: metadata.name
      - name: POD_NAMESPACE
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: metadata.namespace
      - name: POD_IP
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: status.podIP
      - name: HOME
        value: /home
      - name: VLLM_LOGGING_LEVEL
        value: INFO
      - name: HF_HUB_CACHE
        value: /models
      image: ghcr.io/llm-d/llm-d-inference-sim:v0.8.2
      imagePullPolicy: Always
      lifecycle:
        preStop:
          exec:
            command:
            - /bin/sleep
            - "15"
      livenessProbe:
        failureThreshold: 10
        httpGet:
          path: /health
          port: https
          scheme: HTTPS
        periodSeconds: 10
        successThreshold: 1
        timeoutSeconds: 1
      name: main
      ports:
      - containerPort: 8000
        name: https
        protocol: TCP
      readinessProbe:
        failureThreshold: 2
        httpGet:
          path: /ready
          port: https
          scheme: HTTPS
        periodSeconds: 1
        successThreshold: 1
        timeoutSeconds: 1
      resources:
        limits:
          cpu: 500m
          memory: 512Mi
        requests:
          cpu: 100m
          memory: 256Mi
      securityContext:
        allowPrivilegeEscalation: false
        capabilities:
          drop:
          - ALL
        readOnlyRootFilesystem: false
        runAsNonRoot: true
        runAsUser: 1000710000
        seccompProfile:
          type: RuntimeDefault
      startupProbe:
        failureThreshold: 60
        httpGet:
          path: /health
          port: 8000
          scheme: HTTPS
        periodSeconds: 10
        successThreshold: 1
        timeoutSeconds: 1
      terminationMessagePath: /dev/termination-log
      terminationMessagePolicy: FallbackToLogsOnError
      volumeMounts:
      - mountPath: /home
        name: home
      - mountPath: /tmp
        name: tmp-dir
      - mountPath: /dev/shm
        name: dshm
      - mountPath: /models
        name: model-cache
      - mountPath: /var/run/kserve/tls
        name: tls-certs
        readOnly: true
      - mountPath: /var/run/secrets/kubernetes.io/serviceaccount
        name: kube-api-access-hd2zj
        readOnly: true
    dnsPolicy: ClusterFirst
    enableServiceLinks: true
    imagePullSecrets:
    - name: default-dockercfg-p8m6b
    nodeName: ip-10-0-135-24.us-west-1.compute.internal
    preemptionPolicy: PreemptLowerPriority
    priority: 0
    restartPolicy: Always
    schedulerName: default-scheduler
    securityContext:
      fsGroup: 1000710000
      seLinuxOptions:
        level: s0:c27,c4
      seccompProfile:
        type: RuntimeDefault
    serviceAccount: default
    serviceAccountName: default
    terminationGracePeriodSeconds: 60
    tolerations:
    - effect: NoExecute
      key: node.kubernetes.io/not-ready
      operator: Exists
      tolerationSeconds: 300
    - effect: NoExecute
      key: node.kubernetes.io/unreachable
      operator: Exists
      tolerationSeconds: 300
    - effect: NoSchedule
      key: node.kubernetes.io/memory-pressure
      operator: Exists
    volumes:
    - emptyDir: {}
      name: home
    - emptyDir:
        medium: Memory
        sizeLimit: 1Gi
      name: dshm
    - emptyDir: {}
      name: model-cache
    - emptyDir: {}
      name: tmp-dir
    - name: tls-certs
      secret:
        defaultMode: 420
        secretName: e2e-trlp-test-simulated-kserve-self-signed-certs
    - name: kube-api-access-hd2zj
      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
        - configMap:
            items:
            - key: service-ca.crt
              path: service-ca.crt
            name: openshift-service-ca.crt
  status:
    conditions:
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:06Z"
      observedGeneration: 1
      status: "True"
      type: PodReadyToStartContainers
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:04Z"
      observedGeneration: 1
      status: "True"
      type: Initialized
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:14Z"
      observedGeneration: 1
      status: "True"
      type: Ready
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:14Z"
      observedGeneration: 1
      status: "True"
      type: ContainersReady
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:04Z"
      observedGeneration: 1
      status: "True"
      type: PodScheduled
    containerStatuses:
    - allocatedResources:
        cpu: 100m
        memory: 256Mi
      containerID: cri-o://8b1e0df20dabab3d8f504a23acc1faed188aad712159e53b8b8ff525ea25061b
      image: ghcr.io/llm-d/llm-d-inference-sim:v0.8.2
      imageID: ghcr.io/llm-d/llm-d-inference-sim@sha256:bab162bd25e2ed8b15022387cdb223023aeb33be49476af9f0115c0398fb8ff5
      lastState: {}
      name: main
      ready: true
      resources:
        limits:
          cpu: 500m
          memory: 512Mi
        requests:
          cpu: 100m
          memory: 256Mi
      restartCount: 0
      started: true
      state:
        running:
          startedAt: "2026-11-02T12:07:06Z"
      user:
        linux:
          gid: 0
          supplementalGroups:
          - 0
          - 1000710000
          uid: 1000710000
      volumeMounts:
      - mountPath: /home
        name: home
      - mountPath: /tmp
        name: tmp-dir
      - mountPath: /dev/shm
        name: dshm
      - mountPath: /models
        name: model-cache
      - mountPath: /var/run/kserve/tls
        name: tls-certs
        readOnly: true
        recursiveReadOnly: Disabled
      - mountPath: /var/run/secrets/kubernetes.io/serviceaccount
        name: kube-api-access-hd2zj
        readOnly: true
        recursiveReadOnly: Disabled
    hostIP: 10.0.135.24
    hostIPs:
    - ip: 10.0.135.24
    observedGeneration: 1
    phase: Running
    podIP: 10.132.0.49
    podIPs:
    - ip: 10.132.0.49
    qosClass: Burstable
    startTime: "2026-11-02T12:07:04Z"
- apiVersion: v1
  kind: Pod
  metadata:
    annotations:
      k8s.ovn.org/pod-networks: '{"default":{"ip_addresses":["10.132.0.50/23"],"mac_address":"0a:58:0a:84:00:32","gateway_ips":["10.132.0.1"],"routes":[{"dest":"10.132.0.0/14","nextHop":"10.132.0.1"},{"dest":"172.31.0.0/16","nextHop":"10.132.0.1"},{"dest":"169.254.0.5/32","nextHop":"10.132.0.1"},{"dest":"100.64.0.0/16","nextHop":"10.132.0.1"}],"ip_address":"10.132.0.50/23","gateway_ip":"10.132.0.1","role":"primary"}}'
      k8s.v1.cni.cncf.io/network-status: |-
        [{
            "name": "ovn-kubernetes",
            "interface": "eth0",
            "ips": [
                "10.132.0.50"
            ],
            "mac": "0a:58:0a:84:00:32",
            "default": true,
            "dns": {}
        }]
      openshift.io/scc: restricted-v2
      seccomp.security.alpha.kubernetes.io/pod: runtime/default
      security.openshift.io/validated-scc-subject-type: user
    creationTimestamp: "2026-11-02T12:07:07Z"
    generateName: e2e-unconfigured-facebook-opt-125m-simulated-kserve-7658b8f4fc-
    generation: 1
    labels:
      app.kubernetes.io/component: llminferenceservice-workload
      app.kubernetes.io/name: e2e-unconfigured-facebook-opt-125m-simulated
      app.kubernetes.io/part-of: llminferenceservice
      kserve.io/component: workload
      llm-d.ai/role: both
      pod-template-hash: 7658b8f4fc
      topology.kubernetes.io/region: us-west-1
      topology.kubernetes.io/zone: us-west-1a
    managedFields:
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            f:k8s.ovn.org/pod-networks: {}
      manager: ip-10-0-135-24
      operation: Update
      subresource: status
      time: "2026-11-02T12:07:07Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:generateName: {}
          f:labels:
            .: {}
            f:app.kubernetes.io/component: {}
            f:app.kubernetes.io/name: {}
            f:app.kubernetes.io/part-of: {}
            f:kserve.io/component: {}
            f:llm-d.ai/role: {}
            f:pod-template-hash: {}
          f:ownerReferences:
            .: {}
            k:{"uid":"89ad5932-f709-404b-a811-5495fb193886"}: {}
        f:spec:
          f:containers:
            k:{"name":"main"}:
              .: {}
              f:args: {}
              f:command: {}
              f:env:
                .: {}
                k:{"name":"HF_HUB_CACHE"}:
                  .: {}
                  f:name: {}
                  f:value: {}
                k:{"name":"HOME"}:
                  .: {}
                  f:name: {}
                  f:value: {}
                k:{"name":"POD_IP"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"POD_NAME"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"POD_NAMESPACE"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"VLLM_LOGGING_LEVEL"}:
                  .: {}
                  f:name: {}
                  f:value: {}
              f:image: {}
              f:imagePullPolicy: {}
              f:lifecycle:
                .: {}
                f:preStop:
                  .: {}
                  f:exec:
                    .: {}
                    f:command: {}
              f:livenessProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:name: {}
              f:ports:
                .: {}
                k:{"containerPort":8000,"protocol":"TCP"}:
                  .: {}
                  f:containerPort: {}
                  f:name: {}
                  f:protocol: {}
              f:readinessProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:resources:
                .: {}
                f:limits:
                  .: {}
                  f:cpu: {}
                  f:memory: {}
                f:requests:
                  .: {}
                  f:cpu: {}
                  f:memory: {}
              f:securityContext:
                .: {}
                f:allowPrivilegeEscalation: {}
                f:capabilities:
                  .: {}
                  f:drop: {}
                f:readOnlyRootFilesystem: {}
                f:runAsNonRoot: {}
                f:seccompProfile:
                  .: {}
                  f:type: {}
              f:startupProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:terminationMessagePath: {}
              f:terminationMessagePolicy: {}
              f:volumeMounts:
                .: {}
                k:{"mountPath":"/dev/shm"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/home"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/models"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/tmp"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/var/run/kserve/tls"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                  f:readOnly: {}
          f:dnsPolicy: {}
          f:enableServiceLinks: {}
          f:restartPolicy: {}
          f:schedulerName: {}
          f:securityContext: {}
          f:terminationGracePeriodSeconds: {}
          f:volumes:
            .: {}
            k:{"name":"dshm"}:
              .: {}
              f:emptyDir:
                .: {}
                f:medium: {}
                f:sizeLimit: {}
              f:name: {}
            k:{"name":"home"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
            k:{"name":"model-cache"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
            k:{"name":"tls-certs"}:
              .: {}
              f:name: {}
              f:secret:
                .: {}
                f:defaultMode: {}
                f:secretName: {}
            k:{"name":"tmp-dir"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
      manager: kube-controller-manager
      operation: Update
      time: "2026-11-02T12:07:07Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            f:k8s.v1.cni.cncf.io/network-status: {}
      manager: multus-daemon
      operation: Update
      subresource: status
      time: "2026-11-02T12:07:08Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:status:
          f:conditions:
            k:{"type":"ContainersReady"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"Initialized"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"PodReadyToStartContainers"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"PodScheduled"}:
              f:observedGeneration: {}
            k:{"type":"Ready"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
          f:containerStatuses: {}
          f:hostIP: {}
          f:hostIPs: {}
          f:observedGeneration: {}
          f:phase: {}
          f:podIP: {}
          f:podIPs:
            .: {}
            k:{"ip":"10.132.0.50"}:
              .: {}
              f:ip: {}
          f:startTime: {}
      manager: kubelet
      operation: Update
      subresource: status
      time: "2026-11-02T12:07:18Z"
    name: e2e-unconfigured-facebook-opt-125m-simulated-kserve-7658b8k5m6q
    namespace: llm
    ownerReferences:
    - apiVersion: apps/v1
      blockOwnerDeletion: true
      controller: true
      kind: ReplicaSet
      name: e2e-unconfigured-facebook-opt-125m-simulated-kserve-7658b8f4fc
      uid: 89ad5932-f709-404b-a811-5495fb193886
    resourceVersion: "20821"
    uid: ebd62eee-d630-4bab-8d45-b766272f482a
  spec:
    containers:
    - args:
      - --port
      - "8000"
      - --model
      - test/e2e-unconfigured-model
      - --mode
      - random
      - --no-mm-encoder-only
      - --ssl-certfile
      - /var/run/kserve/tls/tls.crt
      - --ssl-keyfile
      - /var/run/kserve/tls/tls.key
      command:
      - /app/llm-d-inference-sim
      env:
      - name: POD_NAME
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: metadata.name
      - name: POD_NAMESPACE
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: metadata.namespace
      - name: POD_IP
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: status.podIP
      - name: HOME
        value: /home
      - name: VLLM_LOGGING_LEVEL
        value: INFO
      - name: HF_HUB_CACHE
        value: /models
      image: ghcr.io/llm-d/llm-d-inference-sim:v0.8.2
      imagePullPolicy: Always
      lifecycle:
        preStop:
          exec:
            command:
            - /bin/sleep
            - "15"
      livenessProbe:
        failureThreshold: 10
        httpGet:
          path: /health
          port: https
          scheme: HTTPS
        periodSeconds: 10
        successThreshold: 1
        timeoutSeconds: 1
      name: main
      ports:
      - containerPort: 8000
        name: https
        protocol: TCP
      readinessProbe:
        failureThreshold: 2
        httpGet:
          path: /ready
          port: https
          scheme: HTTPS
        periodSeconds: 1
        successThreshold: 1
        timeoutSeconds: 1
      resources:
        limits:
          cpu: 500m
          memory: 512Mi
        requests:
          cpu: 100m
          memory: 256Mi
      securityContext:
        allowPrivilegeEscalation: false
        capabilities:
          drop:
          - ALL
        readOnlyRootFilesystem: false
        runAsNonRoot: true
        runAsUser: 1000710000
        seccompProfile:
          type: RuntimeDefault
      startupProbe:
        failureThreshold: 60
        httpGet:
          path: /health
          port: 8000
          scheme: HTTPS
        periodSeconds: 10
        successThreshold: 1
        timeoutSeconds: 1
      terminationMessagePath: /dev/termination-log
      terminationMessagePolicy: FallbackToLogsOnError
      volumeMounts:
      - mountPath: /home
        name: home
      - mountPath: /tmp
        name: tmp-dir
      - mountPath: /dev/shm
        name: dshm
      - mountPath: /models
        name: model-cache
      - mountPath: /var/run/kserve/tls
        name: tls-certs
        readOnly: true
      - mountPath: /var/run/secrets/kubernetes.io/serviceaccount
        name: kube-api-access-z7trn
        readOnly: true
    dnsPolicy: ClusterFirst
    enableServiceLinks: true
    imagePullSecrets:
    - name: default-dockercfg-p8m6b
    nodeName: ip-10-0-135-24.us-west-1.compute.internal
    preemptionPolicy: PreemptLowerPriority
    priority: 0
    restartPolicy: Always
    schedulerName: default-scheduler
    securityContext:
      fsGroup: 1000710000
      seLinuxOptions:
        level: s0:c27,c4
      seccompProfile:
        type: RuntimeDefault
    serviceAccount: default
    serviceAccountName: default
    terminationGracePeriodSeconds: 60
    tolerations:
    - effect: NoExecute
      key: node.kubernetes.io/not-ready
      operator: Exists
      tolerationSeconds: 300
    - effect: NoExecute
      key: node.kubernetes.io/unreachable
      operator: Exists
      tolerationSeconds: 300
    - effect: NoSchedule
      key: node.kubernetes.io/memory-pressure
      operator: Exists
    volumes:
    - emptyDir: {}
      name: home
    - emptyDir:
        medium: Memory
        sizeLimit: 1Gi
      name: dshm
    - emptyDir: {}
      name: model-cache
    - emptyDir: {}
      name: tmp-dir
    - name: tls-certs
      secret:
        defaultMode: 420
        secretName: e2e-unab60ef4d3a239b5143b412cab04acac3-kserve-self-signed-certs
    - name: kube-api-access-z7trn
      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
        - configMap:
            items:
            - key: service-ca.crt
              path: service-ca.crt
            name: openshift-service-ca.crt
  status:
    conditions:
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:09Z"
      observedGeneration: 1
      status: "True"
      type: PodReadyToStartContainers
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:07Z"
      observedGeneration: 1
      status: "True"
      type: Initialized
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:18Z"
      observedGeneration: 1
      status: "True"
      type: Ready
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:18Z"
      observedGeneration: 1
      status: "True"
      type: ContainersReady
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:07Z"
      observedGeneration: 1
      status: "True"
      type: PodScheduled
    containerStatuses:
    - allocatedResources:
        cpu: 100m
        memory: 256Mi
      containerID: cri-o://1e850760eddc13175742587ad38de9b602efc5138a4986628af6d9b73e23bf96
      image: ghcr.io/llm-d/llm-d-inference-sim:v0.8.2
      imageID: ghcr.io/llm-d/llm-d-inference-sim@sha256:bab162bd25e2ed8b15022387cdb223023aeb33be49476af9f0115c0398fb8ff5
      lastState: {}
      name: main
      ready: true
      resources:
        limits:
          cpu: 500m
          memory: 512Mi
        requests:
          cpu: 100m
          memory: 256Mi
      restartCount: 0
      started: true
      state:
        running:
          startedAt: "2026-11-02T12:07:09Z"
      user:
        linux:
          gid: 0
          supplementalGroups:
          - 0
          - 1000710000
          uid: 1000710000
      volumeMounts:
      - mountPath: /home
        name: home
      - mountPath: /tmp
        name: tmp-dir
      - mountPath: /dev/shm
        name: dshm
      - mountPath: /models
        name: model-cache
      - mountPath: /var/run/kserve/tls
        name: tls-certs
        readOnly: true
        recursiveReadOnly: Disabled
      - mountPath: /var/run/secrets/kubernetes.io/serviceaccount
        name: kube-api-access-z7trn
        readOnly: true
        recursiveReadOnly: Disabled
    hostIP: 10.0.135.24
    hostIPs:
    - ip: 10.0.135.24
    observedGeneration: 1
    phase: Running
    podIP: 10.132.0.50
    podIPs:
    - ip: 10.132.0.50
    qosClass: Burstable
    startTime: "2026-11-02T12:07:07Z"
- apiVersion: v1
  kind: Pod
  metadata:
    annotations:
      k8s.ovn.org/pod-networks: '{"default":{"ip_addresses":["10.133.0.38/23"],"mac_address":"0a:58:0a:85:00:26","gateway_ips":["10.133.0.1"],"routes":[{"dest":"10.132.0.0/14","nextHop":"10.133.0.1"},{"dest":"172.31.0.0/16","nextHop":"10.133.0.1"},{"dest":"169.254.0.5/32","nextHop":"10.133.0.1"},{"dest":"100.64.0.0/16","nextHop":"10.133.0.1"}],"ip_address":"10.133.0.38/23","gateway_ip":"10.133.0.1","role":"primary"}}'
      k8s.v1.cni.cncf.io/network-status: |-
        [{
            "name": "ovn-kubernetes",
            "interface": "eth0",
            "ips": [
                "10.133.0.38"
            ],
            "mac": "0a:58:0a:85:00:26",
            "default": true,
            "dns": {}
        }]
      openshift.io/scc: restricted-v2
      seccomp.security.alpha.kubernetes.io/pod: runtime/default
      security.openshift.io/validated-scc-subject-type: user
    creationTimestamp: "2026-11-02T12:07:12Z"
    generateName: echo-simulator-kserve-64b6f8f96c-
    generation: 1
    labels:
      app.kubernetes.io/component: llminferenceservice-workload
      app.kubernetes.io/name: echo-simulator
      app.kubernetes.io/part-of: llminferenceservice
      kserve.io/component: workload
      llm-d.ai/role: both
      pod-template-hash: 64b6f8f96c
      topology.kubernetes.io/region: us-west-1
      topology.kubernetes.io/zone: us-west-1a
    managedFields:
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            f:k8s.ovn.org/pod-networks: {}
      manager: ip-10-0-133-239
      operation: Update
      subresource: status
      time: "2026-11-02T12:07:12Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:generateName: {}
          f:labels:
            .: {}
            f:app.kubernetes.io/component: {}
            f:app.kubernetes.io/name: {}
            f:app.kubernetes.io/part-of: {}
            f:kserve.io/component: {}
            f:llm-d.ai/role: {}
            f:pod-template-hash: {}
          f:ownerReferences:
            .: {}
            k:{"uid":"4745d979-7572-43dc-b404-714aba992ff6"}: {}
        f:spec:
          f:containers:
            k:{"name":"main"}:
              .: {}
              f:args: {}
              f:command: {}
              f:env:
                .: {}
                k:{"name":"HF_HUB_CACHE"}:
                  .: {}
                  f:name: {}
                  f:value: {}
                k:{"name":"HOME"}:
                  .: {}
                  f:name: {}
                  f:value: {}
                k:{"name":"POD_IP"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"POD_NAME"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"POD_NAMESPACE"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"VLLM_LOGGING_LEVEL"}:
                  .: {}
                  f:name: {}
                  f:value: {}
              f:image: {}
              f:imagePullPolicy: {}
              f:lifecycle:
                .: {}
                f:preStop:
                  .: {}
                  f:exec:
                    .: {}
                    f:command: {}
              f:livenessProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:name: {}
              f:ports:
                .: {}
                k:{"containerPort":8000,"protocol":"TCP"}:
                  .: {}
                  f:containerPort: {}
                  f:name: {}
                  f:protocol: {}
              f:readinessProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:resources:
                .: {}
                f:limits:
                  .: {}
                  f:cpu: {}
                  f:memory: {}
                f:requests:
                  .: {}
                  f:cpu: {}
                  f:memory: {}
              f:securityContext:
                .: {}
                f:allowPrivilegeEscalation: {}
                f:capabilities:
                  .: {}
                  f:drop: {}
                f:readOnlyRootFilesystem: {}
                f:runAsNonRoot: {}
                f:seccompProfile:
                  .: {}
                  f:type: {}
              f:startupProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:terminationMessagePath: {}
              f:terminationMessagePolicy: {}
              f:volumeMounts:
                .: {}
                k:{"mountPath":"/dev/shm"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/home"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/models"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/tmp"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/var/run/kserve/tls"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                  f:readOnly: {}
          f:dnsPolicy: {}
          f:enableServiceLinks: {}
          f:restartPolicy: {}
          f:schedulerName: {}
          f:securityContext: {}
          f:terminationGracePeriodSeconds: {}
          f:volumes:
            .: {}
            k:{"name":"dshm"}:
              .: {}
              f:emptyDir:
                .: {}
                f:medium: {}
                f:sizeLimit: {}
              f:name: {}
            k:{"name":"home"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
            k:{"name":"model-cache"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
            k:{"name":"tls-certs"}:
              .: {}
              f:name: {}
              f:secret:
                .: {}
                f:defaultMode: {}
                f:secretName: {}
            k:{"name":"tmp-dir"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
      manager: kube-controller-manager
      operation: Update
      time: "2026-11-02T12:07:12Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            f:k8s.v1.cni.cncf.io/network-status: {}
      manager: multus-daemon
      operation: Update
      subresource: status
      time: "2026-11-02T12:07:12Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:status:
          f:conditions:
            k:{"type":"ContainersReady"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"Initialized"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"PodReadyToStartContainers"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"PodScheduled"}:
              f:observedGeneration: {}
            k:{"type":"Ready"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
          f:containerStatuses: {}
          f:hostIP: {}
          f:hostIPs: {}
          f:observedGeneration: {}
          f:phase: {}
          f:podIP: {}
          f:podIPs:
            .: {}
            k:{"ip":"10.133.0.38"}:
              .: {}
              f:ip: {}
          f:startTime: {}
      manager: kubelet
      operation: Update
      subresource: status
      time: "2026-11-02T12:07:23Z"
    name: echo-simulator-kserve-64b6f8f96c-8bkgg
    namespace: llm
    ownerReferences:
    - apiVersion: apps/v1
      blockOwnerDeletion: true
      controller: true
      kind: ReplicaSet
      name: echo-simulator-kserve-64b6f8f96c
      uid: 4745d979-7572-43dc-b404-714aba992ff6
    resourceVersion: "20982"
    uid: c1851d27-686d-4b50-bbf6-db2e372dd09c
  spec:
    containers:
    - args:
      - --port
      - "8000"
      - --model
      - test/e2e-echo-model
      - --mode
      - echo
      - --max-model-len
      - "131072"
      - --no-mm-encoder-only
      - --ssl-certfile
      - /var/run/kserve/tls/tls.crt
      - --ssl-keyfile
      - /var/run/kserve/tls/tls.key
      command:
      - /app/llm-d-inference-sim
      env:
      - name: POD_NAME
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: metadata.name
      - name: POD_NAMESPACE
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: metadata.namespace
      - name: POD_IP
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: status.podIP
      - name: HOME
        value: /home
      - name: VLLM_LOGGING_LEVEL
        value: INFO
      - name: HF_HUB_CACHE
        value: /models
      image: ghcr.io/llm-d/llm-d-inference-sim:v0.9.0
      imagePullPolicy: Always
      lifecycle:
        preStop:
          exec:
            command:
            - /bin/sleep
            - "15"
      livenessProbe:
        failureThreshold: 10
        httpGet:
          path: /health
          port: https
          scheme: HTTPS
        periodSeconds: 10
        successThreshold: 1
        timeoutSeconds: 1
      name: main
      ports:
      - containerPort: 8000
        name: https
        protocol: TCP
      readinessProbe:
        failureThreshold: 2
        httpGet:
          path: /ready
          port: https
          scheme: HTTPS
        periodSeconds: 1
        successThreshold: 1
        timeoutSeconds: 1
      resources:
        limits:
          cpu: 500m
          memory: 512Mi
        requests:
          cpu: 100m
          memory: 256Mi
      securityContext:
        allowPrivilegeEscalation: false
        capabilities:
          drop:
          - ALL
        readOnlyRootFilesystem: false
        runAsNonRoot: true
        runAsUser: 1000710000
        seccompProfile:
          type: RuntimeDefault
      startupProbe:
        failureThreshold: 60
        httpGet:
          path: /health
          port: 8000
          scheme: HTTPS
        periodSeconds: 10
        successThreshold: 1
        timeoutSeconds: 1
      terminationMessagePath: /dev/termination-log
      terminationMessagePolicy: FallbackToLogsOnError
      volumeMounts:
      - mountPath: /home
        name: home
      - mountPath: /tmp
        name: tmp-dir
      - mountPath: /dev/shm
        name: dshm
      - mountPath: /models
        name: model-cache
      - mountPath: /var/run/kserve/tls
        name: tls-certs
        readOnly: true
      - mountPath: /var/run/secrets/kubernetes.io/serviceaccount
        name: kube-api-access-nsdpj
        readOnly: true
    dnsPolicy: ClusterFirst
    enableServiceLinks: true
    imagePullSecrets:
    - name: default-dockercfg-p8m6b
    nodeName: ip-10-0-133-239.us-west-1.compute.internal
    preemptionPolicy: PreemptLowerPriority
    priority: 0
    restartPolicy: Always
    schedulerName: default-scheduler
    securityContext:
      fsGroup: 1000710000
      seLinuxOptions:
        level: s0:c27,c4
      seccompProfile:
        type: RuntimeDefault
    serviceAccount: default
    serviceAccountName: default
    terminationGracePeriodSeconds: 60
    tolerations:
    - effect: NoExecute
      key: node.kubernetes.io/not-ready
      operator: Exists
      tolerationSeconds: 300
    - effect: NoExecute
      key: node.kubernetes.io/unreachable
      operator: Exists
      tolerationSeconds: 300
    - effect: NoSchedule
      key: node.kubernetes.io/memory-pressure
      operator: Exists
    volumes:
    - emptyDir: {}
      name: home
    - emptyDir:
        medium: Memory
        sizeLimit: 1Gi
      name: dshm
    - emptyDir: {}
      name: model-cache
    - emptyDir: {}
      name: tmp-dir
    - name: tls-certs
      secret:
        defaultMode: 420
        secretName: echo-simulator-kserve-self-signed-certs
    - name: kube-api-access-nsdpj
      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
        - configMap:
            items:
            - key: service-ca.crt
              path: service-ca.crt
            name: openshift-service-ca.crt
  status:
    conditions:
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:16Z"
      observedGeneration: 1
      status: "True"
      type: PodReadyToStartContainers
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:12Z"
      observedGeneration: 1
      status: "True"
      type: Initialized
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:23Z"
      observedGeneration: 1
      status: "True"
      type: Ready
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:23Z"
      observedGeneration: 1
      status: "True"
      type: ContainersReady
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:12Z"
      observedGeneration: 1
      status: "True"
      type: PodScheduled
    containerStatuses:
    - allocatedResources:
        cpu: 100m
        memory: 256Mi
      containerID: cri-o://f8926c6e0f568c11e41b9228430e514aaba3657c2c862c29bd879478087a47d9
      image: ghcr.io/llm-d/llm-d-inference-sim:v0.9.0
      imageID: ghcr.io/llm-d/llm-d-inference-sim@sha256:5f182ada4d1c915d660d19a6e512ac15f70530f9f237a1519940ebd38dfdfc8f
      lastState:
        terminated:
          containerID: cri-o://bd574b767ff8729a6688ae56586c036fa4503c73ca55e0a448d1601d9766d157
          exitCode: 137
          finishedAt: "2026-11-02T12:07:15Z"
          message: "sfer-config\": \"\",\n\t  \"enable-kvcache\": false,\n\t  \"enable-prefix-caching\":
            false,\n\t  \"enable-request-id-headers\": false,\n\t  \"enable-sleep-mode\":
            false,\n\t  \"enforce-eager\": false,\n\t  \"event-batch-size\": 16,\n\t
            \ \"failure-injection-rate\": 0,\n\t  \"failure-types\": null,\n\t  \"fake-metrics\":
            null,\n\t  \"fake-metrics-refresh-interval\": 100000000,\n\t  \"global-cache-hit-threshold\":
            0,\n\t  \"hash-seed\": \"\",\n\t  \"inter-token-latency\": 0,\n\t  \"inter-token-latency-std-dev\":
            0,\n\t  \"kv-cache-size\": 1024,\n\t  \"kv-cache-transfer-latency\": 0,\n\t
            \ \"kv-cache-transfer-latency-std-dev\": 0,\n\t  \"kv-cache-transfer-time-per-token\":
            0,\n\t  \"kv-cache-transfer-time-std-dev\": 0,\n\t  \"latency-calculator\":
            \"\",\n\t  \"log-http\": false,\n\t  \"lora-modules\": null,\n\t  \"max-cpu-loras\":
            1,\n\t  \"max-loras\": 1,\n\t  \"max-model-len\": 131072,\n\t  \"max-num-seqs\":
            5,\n\t  \"max-tool-call-array-param-length\": 5,\n\t  \"max-tool-call-integer-param\":
            100,\n\t  \"max-tool-call-number-param\": 100,\n\t  \"max-waiting-queue-length\":
            1000,\n\t  \"min-tool-call-array-param-length\": 1,\n\t  \"min-tool-call-integer-param\":
            0,\n\t  \"min-tool-call-number-param\": 0,\n\t  \"mm-encoder-only\": false,\n\t
            \ \"mm-processor-kwargs\": \"\",\n\t  \"mm-render-timeout\": 60000000000,\n\t
            \ \"mode\": \"echo\",\n\t  \"model\": \"test/e2e-echo-model\",\n\t  \"object-tool-call-not-required-field-probability\":
            50,\n\t  \"port\": 8000,\n\t  \"prefill-overhead\": 0,\n\t  \"prefill-time-per-token\":
            0,\n\t  \"prefill-time-std-dev\": 0,\n\t  \"render-timeout\": 30000000000,\n\t
            \ \"render-url\": \"http://localhost:8082\",\n\t  \"seed\": 1793621235663327200,\n\t
            \ \"self-signed-certs\": false,\n\t  \"served-model-name\": [\n\t    \"test/e2e-echo-model\"\n\t
            \ ],\n\t  \"ssl-certfile\": \"/var/run/kserve/tls/tls.crt\",\n\t  \"ssl-keyfile\":
            \"/var/run/kserve/tls/tls.key\",\n\t  \"startup-duration\": 0,\n\t  \"time-factor-under-load\":
            1,\n\t  \"time-to-first-token\": 0,\n\t  \"time-to-first-token-std-dev\":
            0,\n\t  \"tool-call-not-required-param-probability\": 50,\n\t  \"zmq-endpoint\":
            \"tcp://127.0.0.1:5557\"\n\t}\n >\nI1102 12:07:15.755900       1 tokenizer.go:53]
            \"Model is not a real HF model, using simulated tokenizer\" model=\"test/e2e-echo-model\"\n"
          reason: OOMKilled
          startedAt: "2026-11-02T12:07:15Z"
      name: main
      ready: true
      resources:
        limits:
          cpu: 500m
          memory: 512Mi
        requests:
          cpu: 100m
          memory: 256Mi
      restartCount: 1
      started: true
      state:
        running:
          startedAt: "2026-11-02T12:07:17Z"
      user:
        linux:
          gid: 0
          supplementalGroups:
          - 0
          - 1000710000
          uid: 1000710000
      volumeMounts:
      - mountPath: /home
        name: home
      - mountPath: /tmp
        name: tmp-dir
      - mountPath: /dev/shm
        name: dshm
      - mountPath: /models
        name: model-cache
      - mountPath: /var/run/kserve/tls
        name: tls-certs
        readOnly: true
        recursiveReadOnly: Disabled
      - mountPath: /var/run/secrets/kubernetes.io/serviceaccount
        name: kube-api-access-nsdpj
        readOnly: true
        recursiveReadOnly: Disabled
    hostIP: 10.0.133.239
    hostIPs:
    - ip: 10.0.133.239
    observedGeneration: 1
    phase: Running
    podIP: 10.133.0.38
    podIPs:
    - ip: 10.133.0.38
    qosClass: Burstable
    startTime: "2026-11-02T12:07:12Z"
- apiVersion: v1
  kind: Pod
  metadata:
    annotations:
      k8s.ovn.org/pod-networks: '{"default":{"ip_addresses":["10.132.0.51/23"],"mac_address":"0a:58:0a:84:00:33","gateway_ips":["10.132.0.1"],"routes":[{"dest":"10.132.0.0/14","nextHop":"10.132.0.1"},{"dest":"172.31.0.0/16","nextHop":"10.132.0.1"},{"dest":"169.254.0.5/32","nextHop":"10.132.0.1"},{"dest":"100.64.0.0/16","nextHop":"10.132.0.1"}],"ip_address":"10.132.0.51/23","gateway_ip":"10.132.0.1","role":"primary"}}'
      k8s.v1.cni.cncf.io/network-status: |-
        [{
            "name": "ovn-kubernetes",
            "interface": "eth0",
            "ips": [
                "10.132.0.51"
            ],
            "mac": "0a:58:0a:84:00:33",
            "default": true,
            "dns": {}
        }]
      openshift.io/scc: restricted-v2
      seccomp.security.alpha.kubernetes.io/pod: runtime/default
      security.openshift.io/validated-scc-subject-type: user
    creationTimestamp: "2026-11-02T12:07:15Z"
    generateName: facebook-opt-125m-simulated-kserve-84b4ff656c-
    generation: 1
    labels:
      app.kubernetes.io/component: llminferenceservice-workload
      app.kubernetes.io/name: facebook-opt-125m-simulated
      app.kubernetes.io/part-of: llminferenceservice
      kserve.io/component: workload
      llm-d.ai/role: both
      pod-template-hash: 84b4ff656c
      topology.kubernetes.io/region: us-west-1
      topology.kubernetes.io/zone: us-west-1a
    managedFields:
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            f:k8s.ovn.org/pod-networks: {}
      manager: ip-10-0-135-24
      operation: Update
      subresource: status
      time: "2026-11-02T12:07:15Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:generateName: {}
          f:labels:
            .: {}
            f:app.kubernetes.io/component: {}
            f:app.kubernetes.io/name: {}
            f:app.kubernetes.io/part-of: {}
            f:kserve.io/component: {}
            f:llm-d.ai/role: {}
            f:pod-template-hash: {}
          f:ownerReferences:
            .: {}
            k:{"uid":"9266b2d5-e521-4eec-9eb6-4527d205ee4e"}: {}
        f:spec:
          f:containers:
            k:{"name":"main"}:
              .: {}
              f:args: {}
              f:command: {}
              f:env:
                .: {}
                k:{"name":"HF_HUB_CACHE"}:
                  .: {}
                  f:name: {}
                  f:value: {}
                k:{"name":"HOME"}:
                  .: {}
                  f:name: {}
                  f:value: {}
                k:{"name":"POD_IP"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"POD_NAME"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"POD_NAMESPACE"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"VLLM_LOGGING_LEVEL"}:
                  .: {}
                  f:name: {}
                  f:value: {}
              f:image: {}
              f:imagePullPolicy: {}
              f:lifecycle:
                .: {}
                f:preStop:
                  .: {}
                  f:exec:
                    .: {}
                    f:command: {}
              f:livenessProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:name: {}
              f:ports:
                .: {}
                k:{"containerPort":8000,"protocol":"TCP"}:
                  .: {}
                  f:containerPort: {}
                  f:name: {}
                  f:protocol: {}
              f:readinessProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:resources:
                .: {}
                f:limits:
                  .: {}
                  f:cpu: {}
                  f:memory: {}
                f:requests:
                  .: {}
                  f:cpu: {}
                  f:memory: {}
              f:securityContext:
                .: {}
                f:allowPrivilegeEscalation: {}
                f:capabilities:
                  .: {}
                  f:drop: {}
                f:readOnlyRootFilesystem: {}
                f:runAsNonRoot: {}
                f:seccompProfile:
                  .: {}
                  f:type: {}
              f:startupProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:terminationMessagePath: {}
              f:terminationMessagePolicy: {}
              f:volumeMounts:
                .: {}
                k:{"mountPath":"/dev/shm"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/home"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/models"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/tmp"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/var/run/kserve/tls"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                  f:readOnly: {}
          f:dnsPolicy: {}
          f:enableServiceLinks: {}
          f:restartPolicy: {}
          f:schedulerName: {}
          f:securityContext: {}
          f:terminationGracePeriodSeconds: {}
          f:volumes:
            .: {}
            k:{"name":"dshm"}:
              .: {}
              f:emptyDir:
                .: {}
                f:medium: {}
                f:sizeLimit: {}
              f:name: {}
            k:{"name":"home"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
            k:{"name":"model-cache"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
            k:{"name":"tls-certs"}:
              .: {}
              f:name: {}
              f:secret:
                .: {}
                f:defaultMode: {}
                f:secretName: {}
            k:{"name":"tmp-dir"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
      manager: kube-controller-manager
      operation: Update
      time: "2026-11-02T12:07:15Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            f:k8s.v1.cni.cncf.io/network-status: {}
      manager: multus-daemon
      operation: Update
      subresource: status
      time: "2026-11-02T12:07:16Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:status:
          f:conditions:
            k:{"type":"ContainersReady"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"Initialized"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"PodReadyToStartContainers"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"PodScheduled"}:
              f:observedGeneration: {}
            k:{"type":"Ready"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
          f:containerStatuses: {}
          f:hostIP: {}
          f:hostIPs: {}
          f:observedGeneration: {}
          f:phase: {}
          f:podIP: {}
          f:podIPs:
            .: {}
            k:{"ip":"10.132.0.51"}:
              .: {}
              f:ip: {}
          f:startTime: {}
      manager: kubelet
      operation: Update
      subresource: status
      time: "2026-11-02T12:07:27Z"
    name: facebook-opt-125m-simulated-kserve-84b4ff656c-mghxh
    namespace: llm
    ownerReferences:
    - apiVersion: apps/v1
      blockOwnerDeletion: true
      controller: true
      kind: ReplicaSet
      name: facebook-opt-125m-simulated-kserve-84b4ff656c
      uid: 9266b2d5-e521-4eec-9eb6-4527d205ee4e
    resourceVersion: "21058"
    uid: 430e2338-5daf-4020-9fc0-5645de2bf124
  spec:
    containers:
    - args:
      - --port
      - "8000"
      - --model
      - facebook/opt-125m
      - --mode
      - random
      - --no-mm-encoder-only
      - --ssl-certfile
      - /var/run/kserve/tls/tls.crt
      - --ssl-keyfile
      - /var/run/kserve/tls/tls.key
      command:
      - /app/llm-d-inference-sim
      env:
      - name: POD_NAME
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: metadata.name
      - name: POD_NAMESPACE
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: metadata.namespace
      - name: POD_IP
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: status.podIP
      - name: HOME
        value: /home
      - name: VLLM_LOGGING_LEVEL
        value: INFO
      - name: HF_HUB_CACHE
        value: /models
      image: ghcr.io/llm-d/llm-d-inference-sim:v0.8.2
      imagePullPolicy: Always
      lifecycle:
        preStop:
          exec:
            command:
            - /bin/sleep
            - "15"
      livenessProbe:
        failureThreshold: 10
        httpGet:
          path: /health
          port: https
          scheme: HTTPS
        periodSeconds: 10
        successThreshold: 1
        timeoutSeconds: 1
      name: main
      ports:
      - containerPort: 8000
        name: https
        protocol: TCP
      readinessProbe:
        failureThreshold: 2
        httpGet:
          path: /ready
          port: https
          scheme: HTTPS
        periodSeconds: 1
        successThreshold: 1
        timeoutSeconds: 1
      resources:
        limits:
          cpu: 500m
          memory: 512Mi
        requests:
          cpu: 100m
          memory: 256Mi
      securityContext:
        allowPrivilegeEscalation: false
        capabilities:
          drop:
          - ALL
        readOnlyRootFilesystem: false
        runAsNonRoot: true
        runAsUser: 1000710000
        seccompProfile:
          type: RuntimeDefault
      startupProbe:
        failureThreshold: 60
        httpGet:
          path: /health
          port: 8000
          scheme: HTTPS
        periodSeconds: 10
        successThreshold: 1
        timeoutSeconds: 1
      terminationMessagePath: /dev/termination-log
      terminationMessagePolicy: FallbackToLogsOnError
      volumeMounts:
      - mountPath: /home
        name: home
      - mountPath: /tmp
        name: tmp-dir
      - mountPath: /dev/shm
        name: dshm
      - mountPath: /models
        name: model-cache
      - mountPath: /var/run/kserve/tls
        name: tls-certs
        readOnly: true
      - mountPath: /var/run/secrets/kubernetes.io/serviceaccount
        name: kube-api-access-btzck
        readOnly: true
    dnsPolicy: ClusterFirst
    enableServiceLinks: true
    imagePullSecrets:
    - name: default-dockercfg-p8m6b
    nodeName: ip-10-0-135-24.us-west-1.compute.internal
    preemptionPolicy: PreemptLowerPriority
    priority: 0
    restartPolicy: Always
    schedulerName: default-scheduler
    securityContext:
      fsGroup: 1000710000
      seLinuxOptions:
        level: s0:c27,c4
      seccompProfile:
        type: RuntimeDefault
    serviceAccount: default
    serviceAccountName: default
    terminationGracePeriodSeconds: 60
    tolerations:
    - effect: NoExecute
      key: node.kubernetes.io/not-ready
      operator: Exists
      tolerationSeconds: 300
    - effect: NoExecute
      key: node.kubernetes.io/unreachable
      operator: Exists
      tolerationSeconds: 300
    - effect: NoSchedule
      key: node.kubernetes.io/memory-pressure
      operator: Exists
    volumes:
    - emptyDir: {}
      name: home
    - emptyDir:
        medium: Memory
        sizeLimit: 1Gi
      name: dshm
    - emptyDir: {}
      name: model-cache
    - emptyDir: {}
      name: tmp-dir
    - name: tls-certs
      secret:
        defaultMode: 420
        secretName: facebook-opt-125m-simulated-kserve-self-signed-certs
    - name: kube-api-access-btzck
      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
        - configMap:
            items:
            - key: service-ca.crt
              path: service-ca.crt
            name: openshift-service-ca.crt
  status:
    conditions:
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:17Z"
      observedGeneration: 1
      status: "True"
      type: PodReadyToStartContainers
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:15Z"
      observedGeneration: 1
      status: "True"
      type: Initialized
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:26Z"
      observedGeneration: 1
      status: "True"
      type: Ready
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:26Z"
      observedGeneration: 1
      status: "True"
      type: ContainersReady
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:15Z"
      observedGeneration: 1
      status: "True"
      type: PodScheduled
    containerStatuses:
    - allocatedResources:
        cpu: 100m
        memory: 256Mi
      containerID: cri-o://2574d386ad49550e3ab4b96c8087579ba860e3c4a03a476481f33490874e5be9
      image: ghcr.io/llm-d/llm-d-inference-sim:v0.8.2
      imageID: ghcr.io/llm-d/llm-d-inference-sim@sha256:bab162bd25e2ed8b15022387cdb223023aeb33be49476af9f0115c0398fb8ff5
      lastState: {}
      name: main
      ready: true
      resources:
        limits:
          cpu: 500m
          memory: 512Mi
        requests:
          cpu: 100m
          memory: 256Mi
      restartCount: 0
      started: true
      state:
        running:
          startedAt: "2026-11-02T12:07:17Z"
      user:
        linux:
          gid: 0
          supplementalGroups:
          - 0
          - 1000710000
          uid: 1000710000
      volumeMounts:
      - mountPath: /home
        name: home
      - mountPath: /tmp
        name: tmp-dir
      - mountPath: /dev/shm
        name: dshm
      - mountPath: /models
        name: model-cache
      - mountPath: /var/run/kserve/tls
        name: tls-certs
        readOnly: true
        recursiveReadOnly: Disabled
      - mountPath: /var/run/secrets/kubernetes.io/serviceaccount
        name: kube-api-access-btzck
        readOnly: true
        recursiveReadOnly: Disabled
    hostIP: 10.0.135.24
    hostIPs:
    - ip: 10.0.135.24
    observedGeneration: 1
    phase: Running
    podIP: 10.132.0.51
    podIPs:
    - ip: 10.132.0.51
    qosClass: Burstable
    startTime: "2026-11-02T12:07:15Z"
- apiVersion: v1
  kind: Pod
  metadata:
    annotations:
      k8s.ovn.org/pod-networks: '{"default":{"ip_addresses":["10.132.0.52/23"],"mac_address":"0a:58:0a:84:00:34","gateway_ips":["10.132.0.1"],"routes":[{"dest":"10.132.0.0/14","nextHop":"10.132.0.1"},{"dest":"172.31.0.0/16","nextHop":"10.132.0.1"},{"dest":"169.254.0.5/32","nextHop":"10.132.0.1"},{"dest":"100.64.0.0/16","nextHop":"10.132.0.1"}],"ip_address":"10.132.0.52/23","gateway_ip":"10.132.0.1","role":"primary"}}'
      k8s.v1.cni.cncf.io/network-status: |-
        [{
            "name": "ovn-kubernetes",
            "interface": "eth0",
            "ips": [
                "10.132.0.52"
            ],
            "mac": "0a:58:0a:84:00:34",
            "default": true,
            "dns": {}
        }]
      openshift.io/scc: restricted-v2
      seccomp.security.alpha.kubernetes.io/pod: runtime/default
      security.openshift.io/validated-scc-subject-type: user
    creationTimestamp: "2026-11-02T12:07:20Z"
    generateName: premium-simulated-simulated-premium-kserve-5f5c4b459f-
    generation: 1
    labels:
      app.kubernetes.io/component: llminferenceservice-workload
      app.kubernetes.io/name: premium-simulated-simulated-premium
      app.kubernetes.io/part-of: llminferenceservice
      kserve.io/component: workload
      llm-d.ai/role: both
      pod-template-hash: 5f5c4b459f
      topology.kubernetes.io/region: us-west-1
      topology.kubernetes.io/zone: us-west-1a
    managedFields:
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            f:k8s.ovn.org/pod-networks: {}
      manager: ip-10-0-135-24
      operation: Update
      subresource: status
      time: "2026-11-02T12:07:20Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:generateName: {}
          f:labels:
            .: {}
            f:app.kubernetes.io/component: {}
            f:app.kubernetes.io/name: {}
            f:app.kubernetes.io/part-of: {}
            f:kserve.io/component: {}
            f:llm-d.ai/role: {}
            f:pod-template-hash: {}
          f:ownerReferences:
            .: {}
            k:{"uid":"dabdfe56-0cfe-439b-a24d-9ceea2e37a44"}: {}
        f:spec:
          f:containers:
            k:{"name":"main"}:
              .: {}
              f:args: {}
              f:command: {}
              f:env:
                .: {}
                k:{"name":"HF_HUB_CACHE"}:
                  .: {}
                  f:name: {}
                  f:value: {}
                k:{"name":"HOME"}:
                  .: {}
                  f:name: {}
                  f:value: {}
                k:{"name":"POD_IP"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"POD_NAME"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"POD_NAMESPACE"}:
                  .: {}
                  f:name: {}
                  f:valueFrom:
                    .: {}
                    f:fieldRef: {}
                k:{"name":"VLLM_LOGGING_LEVEL"}:
                  .: {}
                  f:name: {}
                  f:value: {}
              f:image: {}
              f:imagePullPolicy: {}
              f:lifecycle:
                .: {}
                f:preStop:
                  .: {}
                  f:exec:
                    .: {}
                    f:command: {}
              f:livenessProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:name: {}
              f:ports:
                .: {}
                k:{"containerPort":8000,"protocol":"TCP"}:
                  .: {}
                  f:containerPort: {}
                  f:name: {}
                  f:protocol: {}
              f:readinessProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:resources: {}
              f:securityContext:
                .: {}
                f:allowPrivilegeEscalation: {}
                f:capabilities:
                  .: {}
                  f:drop: {}
                f:readOnlyRootFilesystem: {}
                f:runAsNonRoot: {}
                f:seccompProfile:
                  .: {}
                  f:type: {}
              f:startupProbe:
                .: {}
                f:failureThreshold: {}
                f:httpGet:
                  .: {}
                  f:path: {}
                  f:port: {}
                  f:scheme: {}
                f:periodSeconds: {}
                f:successThreshold: {}
                f:timeoutSeconds: {}
              f:terminationMessagePath: {}
              f:terminationMessagePolicy: {}
              f:volumeMounts:
                .: {}
                k:{"mountPath":"/dev/shm"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/home"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/models"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/tmp"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                k:{"mountPath":"/var/run/kserve/tls"}:
                  .: {}
                  f:mountPath: {}
                  f:name: {}
                  f:readOnly: {}
          f:dnsPolicy: {}
          f:enableServiceLinks: {}
          f:restartPolicy: {}
          f:schedulerName: {}
          f:securityContext: {}
          f:terminationGracePeriodSeconds: {}
          f:volumes:
            .: {}
            k:{"name":"dshm"}:
              .: {}
              f:emptyDir:
                .: {}
                f:medium: {}
                f:sizeLimit: {}
              f:name: {}
            k:{"name":"home"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
            k:{"name":"model-cache"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
            k:{"name":"tls-certs"}:
              .: {}
              f:name: {}
              f:secret:
                .: {}
                f:defaultMode: {}
                f:secretName: {}
            k:{"name":"tmp-dir"}:
              .: {}
              f:emptyDir: {}
              f:name: {}
      manager: kube-controller-manager
      operation: Update
      time: "2026-11-02T12:07:20Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            f:k8s.v1.cni.cncf.io/network-status: {}
      manager: multus-daemon
      operation: Update
      subresource: status
      time: "2026-11-02T12:07:21Z"
    - apiVersion: v1
      fieldsType: FieldsV1
      fieldsV1:
        f:status:
          f:conditions:
            k:{"type":"ContainersReady"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"Initialized"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"PodReadyToStartContainers"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
            k:{"type":"PodScheduled"}:
              f:observedGeneration: {}
            k:{"type":"Ready"}:
              .: {}
              f:lastProbeTime: {}
              f:lastTransitionTime: {}
              f:observedGeneration: {}
              f:status: {}
              f:type: {}
          f:containerStatuses: {}
          f:hostIP: {}
          f:hostIPs: {}
          f:observedGeneration: {}
          f:phase: {}
          f:podIP: {}
          f:podIPs:
            .: {}
            k:{"ip":"10.132.0.52"}:
              .: {}
              f:ip: {}
          f:startTime: {}
      manager: kubelet
      operation: Update
      subresource: status
      time: "2026-11-02T12:19:03Z"
    name: premium-simulated-simulated-premium-kserve-5f5c4b459f-d54rw
    namespace: llm
    ownerReferences:
    - apiVersion: apps/v1
      blockOwnerDeletion: true
      controller: true
      kind: ReplicaSet
      name: premium-simulated-simulated-premium-kserve-5f5c4b459f
      uid: dabdfe56-0cfe-439b-a24d-9ceea2e37a44
    resourceVersion: "41150"
    uid: 242aa9d0-49d5-4981-ad74-0f7b95635e6a
  spec:
    containers:
    - args:
      - --port
      - "8000"
      - --model
      - facebook/opt-125m-premium
      - --mode
      - random
      - --no-mm-encoder-only
      - --ssl-certfile
      - /var/run/kserve/tls/tls.crt
      - --ssl-keyfile
      - /var/run/kserve/tls/tls.key
      command:
      - /app/llm-d-inference-sim
      env:
      - name: POD_NAME
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: metadata.name
      - name: POD_NAMESPACE
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: metadata.namespace
      - name: POD_IP
        valueFrom:
          fieldRef:
            apiVersion: v1
            fieldPath: status.podIP
      - name: HOME
        value: /home
      - name: VLLM_LOGGING_LEVEL
        value: INFO
      - name: HF_HUB_CACHE
        value: /models
      image: ghcr.io/llm-d/llm-d-inference-sim:v0.8.2
      imagePullPolicy: Always
      lifecycle:
        preStop:
          exec:
            command:
            - /bin/sleep
            - "15"
      livenessProbe:
        failureThreshold: 10
        httpGet:
          path: /health
          port: https
          scheme: HTTPS
        periodSeconds: 10
        successThreshold: 1
        timeoutSeconds: 1
      name: main
      ports:
      - containerPort: 8000
        name: https
        protocol: TCP
      readinessProbe:
        failureThreshold: 2
        httpGet:
          path: /ready
          port: https
          scheme: HTTPS
        periodSeconds: 1
        successThreshold: 1
        timeoutSeconds: 1
      resources: {}
      securityContext:
        allowPrivilegeEscalation: false
        capabilities:
          drop:
          - ALL
        readOnlyRootFilesystem: false
        runAsNonRoot: true
        runAsUser: 1000710000
        seccompProfile:
          type: RuntimeDefault
      startupProbe:
        failureThreshold: 60
        httpGet:
          path: /health
          port: 8000
          scheme: HTTPS
        periodSeconds: 10
        successThreshold: 1
        timeoutSeconds: 1
      terminationMessagePath: /dev/termination-log
      terminationMessagePolicy: FallbackToLogsOnError
      volumeMounts:
      - mountPath: /home
        name: home
      - mountPath: /tmp
        name: tmp-dir
      - mountPath: /dev/shm
        name: dshm
      - mountPath: /models
        name: model-cache
      - mountPath: /var/run/kserve/tls
        name: tls-certs
        readOnly: true
      - mountPath: /var/run/secrets/kubernetes.io/serviceaccount
        name: kube-api-access-w55d4
        readOnly: true
    dnsPolicy: ClusterFirst
    enableServiceLinks: true
    imagePullSecrets:
    - name: default-dockercfg-p8m6b
    nodeName: ip-10-0-135-24.us-west-1.compute.internal
    preemptionPolicy: PreemptLowerPriority
    priority: 0
    restartPolicy: Always
    schedulerName: default-scheduler
    securityContext:
      fsGroup: 1000710000
      seLinuxOptions:
        level: s0:c27,c4
      seccompProfile:
        type: RuntimeDefault
    serviceAccount: default
    serviceAccountName: default
    terminationGracePeriodSeconds: 60
    tolerations:
    - effect: NoExecute
      key: node.kubernetes.io/not-ready
      operator: Exists
      tolerationSeconds: 300
    - effect: NoExecute
      key: node.kubernetes.io/unreachable
      operator: Exists
      tolerationSeconds: 300
    volumes:
    - emptyDir: {}
      name: home
    - emptyDir:
        medium: Memory
        sizeLimit: 1Gi
      name: dshm
    - emptyDir: {}
      name: model-cache
    - emptyDir: {}
      name: tmp-dir
    - name: tls-certs
      secret:
        defaultMode: 420
        secretName: premium-simulated-simulated-premium-kserve-self-signed-certs
    - name: kube-api-access-w55d4
      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
        - configMap:
            items:
            - key: service-ca.crt
              path: service-ca.crt
            name: openshift-service-ca.crt
  status:
    conditions:
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:21Z"
      observedGeneration: 1
      status: "True"
      type: PodReadyToStartContainers
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:20Z"
      observedGeneration: 1
      status: "True"
      type: Initialized
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:19:03Z"
      observedGeneration: 1
      status: "True"
      type: Ready
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:19:03Z"
      observedGeneration: 1
      status: "True"
      type: ContainersReady
    - lastProbeTime: null
      lastTransitionTime: "2026-11-02T12:07:20Z"
      observedGeneration: 1
      status: "True"
      type: PodScheduled
    containerStatuses:
    - containerID: cri-o://4c5c41c79c314fba28ca9a5e00d2fef079b74d2834606a94202cd5d0c5173d04
      image: ghcr.io/llm-d/llm-d-inference-sim:v0.8.2
      imageID: ghcr.io/llm-d/llm-d-inference-sim@sha256:bab162bd25e2ed8b15022387cdb223023aeb33be49476af9f0115c0398fb8ff5
      lastState: {}
      name: main
      ready: true
      resources: {}
      restartCount: 0
      started: true
      state:
        running:
          startedAt: "2026-11-02T12:07:21Z"
      user:
        linux:
          gid: 0
          supplementalGroups:
          - 0
          - 1000710000
          uid: 1000710000
      volumeMounts:
      - mountPath: /home
        name: home
      - mountPath: /tmp
        name: tmp-dir
      - mountPath: /dev/shm
        name: dshm
      - mountPath: /models
        name: model-cache
      - mountPath: /var/run/kserve/tls
        name: tls-certs
        readOnly: true
        recursiveReadOnly: Disabled
      - mountPath: /var/run/secrets/kubernetes.io/serviceaccount
        name: kube-api-access-w55d4
        readOnly: true
        recursiveReadOnly: Disabled
    hostIP: 10.0.135.24
    hostIPs:
    - ip: 10.0.135.24
    observedGeneration: 1
    phase: Running
    podIP: 10.132.0.52
    podIPs:
    - ip: 10.132.0.52
    qosClass: BestEffort
    startTime: "2026-11-02T12:07:20Z"
kind: PodList
metadata:
  resourceVersion: "149933"