---
apiVersion: v1
kind: Pod
metadata:
  annotations:
    k8s.ovn.org/pod-networks: '{"default":{"ip_addresses":["10.132.0.45/23"],"mac_address":"0a:58:0a:84:00:2d","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":"10.132.0.45/23","gateway_ip":"10.132.0.1","role":"primary"}}'
    k8s.v1.cni.cncf.io/network-status: |-
      [{
          "name": "ovn-kubernetes",
          "interface": "eth0",
          "ips": [
              "10.132.0.45"
          ],
          "mac": "0a:58:0a:84:00:2d",
          "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: "2026-10-30T01:47:22Z"
  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-east-1
    topology.kubernetes.io/zone: us-east-1a
  managedFields:
  - apiVersion: v1
    fieldsType: FieldsV1
    fieldsV1:
      f:metadata:
        f:annotations:
          f:k8s.ovn.org/pod-networks: {}
    manager: ip-10-0-142-110
    operation: Update
    subresource: status
    time: "2026-10-30T01:47:22Z"
  - 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":"e6149d45-fdab-4e3c-8c80-80b5efd7a88f"}: {}
      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: "2026-10-30T01:47:22Z"
  - 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: "2026-10-30T01:47:23Z"
  - 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":"10.132.0.45"}:
            .: {}
            f:ip: {}
        f:startTime: {}
    manager: kubelet
    operation: Update
    subresource: status
    time: "2026-10-30T01:47:34Z"
  name: echo-simulator-kserve-64b6f8f96c-dw8g7
  namespace: llm
  ownerReferences:
  - apiVersion: apps/v1
    blockOwnerDeletion: true
    controller: true
    kind: ReplicaSet
    name: echo-simulator-kserve-64b6f8f96c
    uid: e6149d45-fdab-4e3c-8c80-80b5efd7a88f
  resourceVersion: "25267"
  uid: dca83999-a27c-4d5b-8ab5-09ccbf01f5a6
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-stmnx
      readOnly: true
  dnsPolicy: ClusterFirst
  enableServiceLinks: true
  imagePullSecrets:
  - name: default-dockercfg-9bctt
  nodeName: ip-10-0-142-110.ec2.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-stmnx
    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: "2026-10-30T01:47:25Z"
    observedGeneration: 1
    status: "True"
    type: PodReadyToStartContainers
  - lastProbeTime: null
    lastTransitionTime: "2026-10-30T01:47:22Z"
    observedGeneration: 1
    status: "True"
    type: Initialized
  - lastProbeTime: null
    lastTransitionTime: "2026-10-30T01:47:34Z"
    observedGeneration: 1
    status: "True"
    type: Ready
  - lastProbeTime: null
    lastTransitionTime: "2026-10-30T01:47:34Z"
    observedGeneration: 1
    status: "True"
    type: ContainersReady
  - lastProbeTime: null
    lastTransitionTime: "2026-10-30T01:47:22Z"
    observedGeneration: 1
    status: "True"
    type: PodScheduled
  containerStatuses:
  - allocatedResources:
      cpu: 100m
      memory: 256Mi
    containerID: cri-o://6531d4c8636e2bcbe6427119cd7df40c0cbc464f0a35feee4a7bad1c4aa27a75
    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://490b55bde8949054fb4d6f3d6d847a6f0a07ca20a67bde5470765a99be73eee1
        exitCode: 137
        finishedAt: "2026-10-30T01:47:25Z"
        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\": 1793324845076985900,\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 >\nI1030 01:47:25.110037       1 tokenizer.go:53]
          \"Model is not a real HF model, using simulated tokenizer\" model=\"test/e2e-echo-model\"\n"
        reason: OOMKilled
        startedAt: "2026-10-30T01:47:25Z"
    name: main
    ready: true
    resources:
      limits:
        cpu: 500m
        memory: 512Mi
      requests:
        cpu: 100m
        memory: 256Mi
    restartCount: 1
    started: true
    state:
      running:
        startedAt: "2026-10-30T01:47:26Z"
    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-stmnx
      readOnly: true
      recursiveReadOnly: Disabled
  hostIP: 10.0.142.110
  hostIPs:
  - ip: 10.0.142.110
  observedGeneration: 1
  phase: Running
  podIP: 10.132.0.45
  podIPs:
  - ip: 10.132.0.45
  qosClass: Burstable
  startTime: "2026-10-30T01:47:22Z"