---
apiVersion: apps/v1
items:
- apiVersion: apps/v1
  kind: Deployment
  metadata:
    annotations:
      deployment.kubernetes.io/revision: "1"
    creationTimestamp: "2026-10-30T00:13:12Z"
    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
    managedFields:
    - apiVersion: apps/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          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:ownerReferences:
            .: {}
            k:{"uid":"a7d70b76-6096-4113-8ee6-1d33bfe7fe54"}: {}
        f:spec:
          f:progressDeadlineSeconds: {}
          f:replicas: {}
          f:revisionHistoryLimit: {}
          f:selector: {}
          f:strategy:
            f:rollingUpdate:
              .: {}
              f:maxSurge: {}
              f:maxUnavailable: {}
            f:type: {}
          f:template:
            f:metadata:
              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: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: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: manager
      operation: Update
      time: "2026-10-30T00:13:12Z"
    - apiVersion: apps/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            .: {}
            f:deployment.kubernetes.io/revision: {}
        f:status:
          f:availableReplicas: {}
          f:conditions:
            .: {}
            k:{"type":"Available"}:
              .: {}
              f:lastTransitionTime: {}
              f:lastUpdateTime: {}
              f:message: {}
              f:reason: {}
              f:status: {}
              f:type: {}
            k:{"type":"Progressing"}:
              .: {}
              f:lastTransitionTime: {}
              f:lastUpdateTime: {}
              f:message: {}
              f:reason: {}
              f:status: {}
              f:type: {}
          f:observedGeneration: {}
          f:readyReplicas: {}
          f:replicas: {}
          f:terminatingReplicas: {}
          f:updatedReplicas: {}
      manager: kube-controller-manager
      operation: Update
      subresource: status
      time: "2026-10-30T00:13:23Z"
    name: e2e-distinct-2-simulated-kserve
    namespace: llm
    ownerReferences:
    - apiVersion: serving.kserve.io/v1alpha2
      blockOwnerDeletion: true
      controller: true
      kind: LLMInferenceService
      name: e2e-distinct-2-simulated
      uid: a7d70b76-6096-4113-8ee6-1d33bfe7fe54
    resourceVersion: "19849"
    uid: 14609740-0b06-4bb3-b583-79432af82002
  spec:
    progressDeadlineSeconds: 600
    replicas: 1
    revisionHistoryLimit: 10
    selector:
      matchLabels:
        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
    strategy:
      rollingUpdate:
        maxSurge: 25%
        maxUnavailable: 25%
      type: RollingUpdate
    template:
      metadata:
        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
      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
            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
        dnsPolicy: ClusterFirst
        restartPolicy: Always
        schedulerName: default-scheduler
        securityContext: {}
        terminationGracePeriodSeconds: 60
        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
  status:
    availableReplicas: 1
    conditions:
    - lastTransitionTime: "2026-10-30T00:13:23Z"
      lastUpdateTime: "2026-10-30T00:13:23Z"
      message: Deployment has minimum availability.
      reason: MinimumReplicasAvailable
      status: "True"
      type: Available
    - lastTransitionTime: "2026-10-30T00:13:12Z"
      lastUpdateTime: "2026-10-30T00:13:23Z"
      message: ReplicaSet "e2e-distinct-2-simulated-kserve-cbd4f5586" has successfully
        progressed.
      reason: NewReplicaSetAvailable
      status: "True"
      type: Progressing
    observedGeneration: 1
    readyReplicas: 1
    replicas: 1
    terminatingReplicas: 0
    updatedReplicas: 1
- apiVersion: apps/v1
  kind: Deployment
  metadata:
    annotations:
      deployment.kubernetes.io/revision: "1"
    creationTimestamp: "2026-10-30T00:13:16Z"
    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
    managedFields:
    - apiVersion: apps/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          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:ownerReferences:
            .: {}
            k:{"uid":"7ede888b-6e78-4688-88e7-525cdfaa721d"}: {}
        f:spec:
          f:progressDeadlineSeconds: {}
          f:replicas: {}
          f:revisionHistoryLimit: {}
          f:selector: {}
          f:strategy:
            f:rollingUpdate:
              .: {}
              f:maxSurge: {}
              f:maxUnavailable: {}
            f:type: {}
          f:template:
            f:metadata:
              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: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: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: manager
      operation: Update
      time: "2026-10-30T00:13:16Z"
    - apiVersion: apps/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            .: {}
            f:deployment.kubernetes.io/revision: {}
        f:status:
          f:availableReplicas: {}
          f:conditions:
            .: {}
            k:{"type":"Available"}:
              .: {}
              f:lastTransitionTime: {}
              f:lastUpdateTime: {}
              f:message: {}
              f:reason: {}
              f:status: {}
              f:type: {}
            k:{"type":"Progressing"}:
              .: {}
              f:lastTransitionTime: {}
              f:lastUpdateTime: {}
              f:message: {}
              f:reason: {}
              f:status: {}
              f:type: {}
          f:observedGeneration: {}
          f:readyReplicas: {}
          f:replicas: {}
          f:terminatingReplicas: {}
          f:updatedReplicas: {}
      manager: kube-controller-manager
      operation: Update
      subresource: status
      time: "2026-10-30T00:13:27Z"
    name: e2e-distinct-simulated-kserve
    namespace: llm
    ownerReferences:
    - apiVersion: serving.kserve.io/v1alpha2
      blockOwnerDeletion: true
      controller: true
      kind: LLMInferenceService
      name: e2e-distinct-simulated
      uid: 7ede888b-6e78-4688-88e7-525cdfaa721d
    resourceVersion: "20051"
    uid: 4ca21463-6b11-41a0-8154-8327d7120d98
  spec:
    progressDeadlineSeconds: 600
    replicas: 1
    revisionHistoryLimit: 10
    selector:
      matchLabels:
        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
    strategy:
      rollingUpdate:
        maxSurge: 25%
        maxUnavailable: 25%
      type: RollingUpdate
    template:
      metadata:
        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
      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
            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
        dnsPolicy: ClusterFirst
        restartPolicy: Always
        schedulerName: default-scheduler
        securityContext: {}
        terminationGracePeriodSeconds: 60
        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
  status:
    availableReplicas: 1
    conditions:
    - lastTransitionTime: "2026-10-30T00:13:27Z"
      lastUpdateTime: "2026-10-30T00:13:27Z"
      message: Deployment has minimum availability.
      reason: MinimumReplicasAvailable
      status: "True"
      type: Available
    - lastTransitionTime: "2026-10-30T00:13:16Z"
      lastUpdateTime: "2026-10-30T00:13:27Z"
      message: ReplicaSet "e2e-distinct-simulated-kserve-7b998b79d9" has successfully
        progressed.
      reason: NewReplicaSetAvailable
      status: "True"
      type: Progressing
    observedGeneration: 1
    readyReplicas: 1
    replicas: 1
    terminatingReplicas: 0
    updatedReplicas: 1
- apiVersion: apps/v1
  kind: Deployment
  metadata:
    annotations:
      deployment.kubernetes.io/revision: "1"
    creationTimestamp: "2026-10-30T00:13:19Z"
    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
    managedFields:
    - apiVersion: apps/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          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:ownerReferences:
            .: {}
            k:{"uid":"54157f17-7076-47ed-a0e9-8be121b931c3"}: {}
        f:spec:
          f:progressDeadlineSeconds: {}
          f:replicas: {}
          f:revisionHistoryLimit: {}
          f:selector: {}
          f:strategy:
            f:rollingUpdate:
              .: {}
              f:maxSurge: {}
              f:maxUnavailable: {}
            f:type: {}
          f:template:
            f:metadata:
              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: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: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: manager
      operation: Update
      time: "2026-10-30T00:13:19Z"
    - apiVersion: apps/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            .: {}
            f:deployment.kubernetes.io/revision: {}
        f:status:
          f:availableReplicas: {}
          f:conditions:
            .: {}
            k:{"type":"Available"}:
              .: {}
              f:lastTransitionTime: {}
              f:lastUpdateTime: {}
              f:message: {}
              f:reason: {}
              f:status: {}
              f:type: {}
            k:{"type":"Progressing"}:
              .: {}
              f:lastTransitionTime: {}
              f:lastUpdateTime: {}
              f:message: {}
              f:reason: {}
              f:status: {}
              f:type: {}
          f:observedGeneration: {}
          f:readyReplicas: {}
          f:replicas: {}
          f:terminatingReplicas: {}
          f:updatedReplicas: {}
      manager: kube-controller-manager
      operation: Update
      subresource: status
      time: "2026-10-30T00:13:30Z"
    name: e2e-embedding-simulated-kserve
    namespace: llm
    ownerReferences:
    - apiVersion: serving.kserve.io/v1alpha2
      blockOwnerDeletion: true
      controller: true
      kind: LLMInferenceService
      name: e2e-embedding-simulated
      uid: 54157f17-7076-47ed-a0e9-8be121b931c3
    resourceVersion: "20235"
    uid: 1e561630-f464-4eec-a261-b19f35ad07f0
  spec:
    progressDeadlineSeconds: 600
    replicas: 1
    revisionHistoryLimit: 10
    selector:
      matchLabels:
        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
    strategy:
      rollingUpdate:
        maxSurge: 25%
        maxUnavailable: 25%
      type: RollingUpdate
    template:
      metadata:
        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
      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
            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
        dnsPolicy: ClusterFirst
        restartPolicy: Always
        schedulerName: default-scheduler
        securityContext: {}
        terminationGracePeriodSeconds: 60
        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
  status:
    availableReplicas: 1
    conditions:
    - lastTransitionTime: "2026-10-30T00:13:30Z"
      lastUpdateTime: "2026-10-30T00:13:30Z"
      message: Deployment has minimum availability.
      reason: MinimumReplicasAvailable
      status: "True"
      type: Available
    - lastTransitionTime: "2026-10-30T00:13:19Z"
      lastUpdateTime: "2026-10-30T00:13:30Z"
      message: ReplicaSet "e2e-embedding-simulated-kserve-67bf594cd6" has successfully
        progressed.
      reason: NewReplicaSetAvailable
      status: "True"
      type: Progressing
    observedGeneration: 1
    readyReplicas: 1
    replicas: 1
    terminatingReplicas: 0
    updatedReplicas: 1
- apiVersion: apps/v1
  kind: Deployment
  metadata:
    annotations:
      deployment.kubernetes.io/revision: "1"
    creationTimestamp: "2026-10-30T00:13:21Z"
    generation: 1
    labels:
      app.kubernetes.io/component: llminferenceservice-workload
      app.kubernetes.io/name: e2e-external-token-limit-simulated
      app.kubernetes.io/part-of: llminferenceservice
      kserve.io/component: workload
      llm-d.ai/role: both
    managedFields:
    - apiVersion: apps/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          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:ownerReferences:
            .: {}
            k:{"uid":"40a3041c-94e2-4c90-ad45-f1ed12233c1a"}: {}
        f:spec:
          f:progressDeadlineSeconds: {}
          f:replicas: {}
          f:revisionHistoryLimit: {}
          f:selector: {}
          f:strategy:
            f:rollingUpdate:
              .: {}
              f:maxSurge: {}
              f:maxUnavailable: {}
            f:type: {}
          f:template:
            f:metadata:
              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: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: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: manager
      operation: Update
      time: "2026-10-30T00:13:20Z"
    - apiVersion: apps/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            .: {}
            f:deployment.kubernetes.io/revision: {}
        f:status:
          f:availableReplicas: {}
          f:conditions:
            .: {}
            k:{"type":"Available"}:
              .: {}
              f:lastTransitionTime: {}
              f:lastUpdateTime: {}
              f:message: {}
              f:reason: {}
              f:status: {}
              f:type: {}
            k:{"type":"Progressing"}:
              .: {}
              f:lastTransitionTime: {}
              f:lastUpdateTime: {}
              f:message: {}
              f:reason: {}
              f:status: {}
              f:type: {}
          f:observedGeneration: {}
          f:readyReplicas: {}
          f:replicas: {}
          f:terminatingReplicas: {}
          f:updatedReplicas: {}
      manager: kube-controller-manager
      operation: Update
      subresource: status
      time: "2026-10-30T00:13:31Z"
    name: e2e-external-token-limit-simulated-kserve
    namespace: llm
    ownerReferences:
    - apiVersion: serving.kserve.io/v1alpha2
      blockOwnerDeletion: true
      controller: true
      kind: LLMInferenceService
      name: e2e-external-token-limit-simulated
      uid: 40a3041c-94e2-4c90-ad45-f1ed12233c1a
    resourceVersion: "20296"
    uid: 16e8b8a9-e062-40fe-bc3a-ce33a2da5506
  spec:
    progressDeadlineSeconds: 600
    replicas: 1
    revisionHistoryLimit: 10
    selector:
      matchLabels:
        app.kubernetes.io/component: llminferenceservice-workload
        app.kubernetes.io/name: e2e-external-token-limit-simulated
        app.kubernetes.io/part-of: llminferenceservice
        kserve.io/component: workload
        llm-d.ai/role: both
    strategy:
      rollingUpdate:
        maxSurge: 25%
        maxUnavailable: 25%
      type: RollingUpdate
    template:
      metadata:
        labels:
          app.kubernetes.io/component: llminferenceservice-workload
          app.kubernetes.io/name: e2e-external-token-limit-simulated
          app.kubernetes.io/part-of: llminferenceservice
          kserve.io/component: workload
          llm-d.ai/role: both
      spec:
        containers:
        - args:
          - --port
          - "8000"
          - --model
          - e2e-external-token-limit-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
            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
        dnsPolicy: ClusterFirst
        restartPolicy: Always
        schedulerName: default-scheduler
        securityContext: {}
        terminationGracePeriodSeconds: 60
        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-external-token-limit-simulated-kserve-self-signed-certs
  status:
    availableReplicas: 1
    conditions:
    - lastTransitionTime: "2026-10-30T00:13:31Z"
      lastUpdateTime: "2026-10-30T00:13:31Z"
      message: Deployment has minimum availability.
      reason: MinimumReplicasAvailable
      status: "True"
      type: Available
    - lastTransitionTime: "2026-10-30T00:13:21Z"
      lastUpdateTime: "2026-10-30T00:13:31Z"
      message: ReplicaSet "e2e-external-token-limit-simulated-kserve-65fc98cd9c" has
        successfully progressed.
      reason: NewReplicaSetAvailable
      status: "True"
      type: Progressing
    observedGeneration: 1
    readyReplicas: 1
    replicas: 1
    terminatingReplicas: 0
    updatedReplicas: 1
- apiVersion: apps/v1
  kind: Deployment
  metadata:
    annotations:
      deployment.kubernetes.io/revision: "1"
    creationTimestamp: "2026-10-30T00:13:22Z"
    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
    managedFields:
    - apiVersion: apps/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          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:ownerReferences:
            .: {}
            k:{"uid":"c5dc13a4-2f11-46db-b21a-611d9ddaa0c0"}: {}
        f:spec:
          f:progressDeadlineSeconds: {}
          f:replicas: {}
          f:revisionHistoryLimit: {}
          f:selector: {}
          f:strategy:
            f:rollingUpdate:
              .: {}
              f:maxSurge: {}
              f:maxUnavailable: {}
            f:type: {}
          f:template:
            f:metadata:
              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: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: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: manager
      operation: Update
      time: "2026-10-30T00:13:22Z"
    - apiVersion: apps/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            .: {}
            f:deployment.kubernetes.io/revision: {}
        f:status:
          f:availableReplicas: {}
          f:conditions:
            .: {}
            k:{"type":"Available"}:
              .: {}
              f:lastTransitionTime: {}
              f:lastUpdateTime: {}
              f:message: {}
              f:reason: {}
              f:status: {}
              f:type: {}
            k:{"type":"Progressing"}:
              .: {}
              f:lastTransitionTime: {}
              f:lastUpdateTime: {}
              f:message: {}
              f:reason: {}
              f:status: {}
              f:type: {}
          f:observedGeneration: {}
          f:readyReplicas: {}
          f:replicas: {}
          f:terminatingReplicas: {}
          f:updatedReplicas: {}
      manager: kube-controller-manager
      operation: Update
      subresource: status
      time: "2026-10-30T00:13:33Z"
    name: e2e-trlp-test-simulated-kserve
    namespace: llm
    ownerReferences:
    - apiVersion: serving.kserve.io/v1alpha2
      blockOwnerDeletion: true
      controller: true
      kind: LLMInferenceService
      name: e2e-trlp-test-simulated
      uid: c5dc13a4-2f11-46db-b21a-611d9ddaa0c0
    resourceVersion: "20414"
    uid: ada967a9-112b-4812-bce9-39a736bbb57a
  spec:
    progressDeadlineSeconds: 600
    replicas: 1
    revisionHistoryLimit: 10
    selector:
      matchLabels:
        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
    strategy:
      rollingUpdate:
        maxSurge: 25%
        maxUnavailable: 25%
      type: RollingUpdate
    template:
      metadata:
        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
      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
            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
        dnsPolicy: ClusterFirst
        restartPolicy: Always
        schedulerName: default-scheduler
        securityContext: {}
        terminationGracePeriodSeconds: 60
        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
  status:
    availableReplicas: 1
    conditions:
    - lastTransitionTime: "2026-10-30T00:13:33Z"
      lastUpdateTime: "2026-10-30T00:13:33Z"
      message: Deployment has minimum availability.
      reason: MinimumReplicasAvailable
      status: "True"
      type: Available
    - lastTransitionTime: "2026-10-30T00:13:22Z"
      lastUpdateTime: "2026-10-30T00:13:33Z"
      message: ReplicaSet "e2e-trlp-test-simulated-kserve-69b588897d" has successfully
        progressed.
      reason: NewReplicaSetAvailable
      status: "True"
      type: Progressing
    observedGeneration: 1
    readyReplicas: 1
    replicas: 1
    terminatingReplicas: 0
    updatedReplicas: 1
- apiVersion: apps/v1
  kind: Deployment
  metadata:
    annotations:
      deployment.kubernetes.io/revision: "1"
    creationTimestamp: "2026-10-30T00:13:25Z"
    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
    managedFields:
    - apiVersion: apps/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          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:ownerReferences:
            .: {}
            k:{"uid":"e3bb26be-e39b-4c77-9544-414dcff9c1a9"}: {}
        f:spec:
          f:progressDeadlineSeconds: {}
          f:replicas: {}
          f:revisionHistoryLimit: {}
          f:selector: {}
          f:strategy:
            f:rollingUpdate:
              .: {}
              f:maxSurge: {}
              f:maxUnavailable: {}
            f:type: {}
          f:template:
            f:metadata:
              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: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: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: manager
      operation: Update
      time: "2026-10-30T00:13:25Z"
    - apiVersion: apps/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            .: {}
            f:deployment.kubernetes.io/revision: {}
        f:status:
          f:availableReplicas: {}
          f:conditions:
            .: {}
            k:{"type":"Available"}:
              .: {}
              f:lastTransitionTime: {}
              f:lastUpdateTime: {}
              f:message: {}
              f:reason: {}
              f:status: {}
              f:type: {}
            k:{"type":"Progressing"}:
              .: {}
              f:lastTransitionTime: {}
              f:lastUpdateTime: {}
              f:message: {}
              f:reason: {}
              f:status: {}
              f:type: {}
          f:observedGeneration: {}
          f:readyReplicas: {}
          f:replicas: {}
          f:terminatingReplicas: {}
          f:updatedReplicas: {}
      manager: kube-controller-manager
      operation: Update
      subresource: status
      time: "2026-10-30T00:13:36Z"
    name: e2e-unconfigured-facebook-opt-125m-simulated-kserve
    namespace: llm
    ownerReferences:
    - apiVersion: serving.kserve.io/v1alpha2
      blockOwnerDeletion: true
      controller: true
      kind: LLMInferenceService
      name: e2e-unconfigured-facebook-opt-125m-simulated
      uid: e3bb26be-e39b-4c77-9544-414dcff9c1a9
    resourceVersion: "20519"
    uid: 5e20042f-5a40-4088-9f55-05500f80ce49
  spec:
    progressDeadlineSeconds: 600
    replicas: 1
    revisionHistoryLimit: 10
    selector:
      matchLabels:
        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
    strategy:
      rollingUpdate:
        maxSurge: 25%
        maxUnavailable: 25%
      type: RollingUpdate
    template:
      metadata:
        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
      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
            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
        dnsPolicy: ClusterFirst
        restartPolicy: Always
        schedulerName: default-scheduler
        securityContext: {}
        terminationGracePeriodSeconds: 60
        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
  status:
    availableReplicas: 1
    conditions:
    - lastTransitionTime: "2026-10-30T00:13:36Z"
      lastUpdateTime: "2026-10-30T00:13:36Z"
      message: Deployment has minimum availability.
      reason: MinimumReplicasAvailable
      status: "True"
      type: Available
    - lastTransitionTime: "2026-10-30T00:13:25Z"
      lastUpdateTime: "2026-10-30T00:13:36Z"
      message: ReplicaSet "e2e-unconfigured-facebook-opt-125m-simulated-kserve-7658b8f4fc"
        has successfully progressed.
      reason: NewReplicaSetAvailable
      status: "True"
      type: Progressing
    observedGeneration: 1
    readyReplicas: 1
    replicas: 1
    terminatingReplicas: 0
    updatedReplicas: 1
- apiVersion: apps/v1
  kind: Deployment
  metadata:
    annotations:
      deployment.kubernetes.io/revision: "1"
    creationTimestamp: "2026-10-30T00:13:27Z"
    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
    managedFields:
    - apiVersion: apps/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          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:ownerReferences:
            .: {}
            k:{"uid":"1f6e723a-e30e-4724-ab36-cc01169f35b9"}: {}
        f:spec:
          f:progressDeadlineSeconds: {}
          f:replicas: {}
          f:revisionHistoryLimit: {}
          f:selector: {}
          f:strategy:
            f:rollingUpdate:
              .: {}
              f:maxSurge: {}
              f:maxUnavailable: {}
            f:type: {}
          f:template:
            f:metadata:
              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: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: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: manager
      operation: Update
      time: "2026-10-30T00:13:27Z"
    - apiVersion: apps/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            .: {}
            f:deployment.kubernetes.io/revision: {}
        f:status:
          f:availableReplicas: {}
          f:conditions:
            .: {}
            k:{"type":"Available"}:
              .: {}
              f:lastTransitionTime: {}
              f:lastUpdateTime: {}
              f:message: {}
              f:reason: {}
              f:status: {}
              f:type: {}
            k:{"type":"Progressing"}:
              .: {}
              f:lastTransitionTime: {}
              f:lastUpdateTime: {}
              f:message: {}
              f:reason: {}
              f:status: {}
              f:type: {}
          f:observedGeneration: {}
          f:readyReplicas: {}
          f:replicas: {}
          f:terminatingReplicas: {}
          f:updatedReplicas: {}
      manager: kube-controller-manager
      operation: Update
      subresource: status
      time: "2026-10-30T00:13:38Z"
    name: echo-simulator-kserve
    namespace: llm
    ownerReferences:
    - apiVersion: serving.kserve.io/v1alpha2
      blockOwnerDeletion: true
      controller: true
      kind: LLMInferenceService
      name: echo-simulator
      uid: 1f6e723a-e30e-4724-ab36-cc01169f35b9
    resourceVersion: "20554"
    uid: a172ab2b-b67b-49e6-a979-5906fb3ce848
  spec:
    progressDeadlineSeconds: 600
    replicas: 1
    revisionHistoryLimit: 10
    selector:
      matchLabels:
        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
    strategy:
      rollingUpdate:
        maxSurge: 25%
        maxUnavailable: 25%
      type: RollingUpdate
    template:
      metadata:
        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
      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
            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
        dnsPolicy: ClusterFirst
        restartPolicy: Always
        schedulerName: default-scheduler
        securityContext: {}
        terminationGracePeriodSeconds: 60
        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
  status:
    availableReplicas: 1
    conditions:
    - lastTransitionTime: "2026-10-30T00:13:38Z"
      lastUpdateTime: "2026-10-30T00:13:38Z"
      message: Deployment has minimum availability.
      reason: MinimumReplicasAvailable
      status: "True"
      type: Available
    - lastTransitionTime: "2026-10-30T00:13:27Z"
      lastUpdateTime: "2026-10-30T00:13:38Z"
      message: ReplicaSet "echo-simulator-kserve-64b6f8f96c" has successfully progressed.
      reason: NewReplicaSetAvailable
      status: "True"
      type: Progressing
    observedGeneration: 1
    readyReplicas: 1
    replicas: 1
    terminatingReplicas: 0
    updatedReplicas: 1
- apiVersion: apps/v1
  kind: Deployment
  metadata:
    annotations:
      deployment.kubernetes.io/revision: "1"
    creationTimestamp: "2026-10-30T00:13:29Z"
    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
    managedFields:
    - apiVersion: apps/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          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:ownerReferences:
            .: {}
            k:{"uid":"81716afe-8073-4dc9-b9d2-cec182309a5a"}: {}
        f:spec:
          f:progressDeadlineSeconds: {}
          f:replicas: {}
          f:revisionHistoryLimit: {}
          f:selector: {}
          f:strategy:
            f:rollingUpdate:
              .: {}
              f:maxSurge: {}
              f:maxUnavailable: {}
            f:type: {}
          f:template:
            f:metadata:
              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: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: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: manager
      operation: Update
      time: "2026-10-30T00:13:29Z"
    - apiVersion: apps/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            .: {}
            f:deployment.kubernetes.io/revision: {}
        f:status:
          f:availableReplicas: {}
          f:conditions:
            .: {}
            k:{"type":"Available"}:
              .: {}
              f:lastTransitionTime: {}
              f:lastUpdateTime: {}
              f:message: {}
              f:reason: {}
              f:status: {}
              f:type: {}
            k:{"type":"Progressing"}:
              .: {}
              f:lastTransitionTime: {}
              f:lastUpdateTime: {}
              f:message: {}
              f:reason: {}
              f:status: {}
              f:type: {}
          f:observedGeneration: {}
          f:readyReplicas: {}
          f:replicas: {}
          f:terminatingReplicas: {}
          f:updatedReplicas: {}
      manager: kube-controller-manager
      operation: Update
      subresource: status
      time: "2026-10-30T00:13:40Z"
    name: facebook-opt-125m-simulated-kserve
    namespace: llm
    ownerReferences:
    - apiVersion: serving.kserve.io/v1alpha2
      blockOwnerDeletion: true
      controller: true
      kind: LLMInferenceService
      name: facebook-opt-125m-simulated
      uid: 81716afe-8073-4dc9-b9d2-cec182309a5a
    resourceVersion: "20605"
    uid: 9ae8342a-c1c0-4f2d-9af5-09303fc823b5
  spec:
    progressDeadlineSeconds: 600
    replicas: 1
    revisionHistoryLimit: 10
    selector:
      matchLabels:
        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
    strategy:
      rollingUpdate:
        maxSurge: 25%
        maxUnavailable: 25%
      type: RollingUpdate
    template:
      metadata:
        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
      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
            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
        dnsPolicy: ClusterFirst
        restartPolicy: Always
        schedulerName: default-scheduler
        securityContext: {}
        terminationGracePeriodSeconds: 60
        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
  status:
    availableReplicas: 1
    conditions:
    - lastTransitionTime: "2026-10-30T00:13:40Z"
      lastUpdateTime: "2026-10-30T00:13:40Z"
      message: Deployment has minimum availability.
      reason: MinimumReplicasAvailable
      status: "True"
      type: Available
    - lastTransitionTime: "2026-10-30T00:13:29Z"
      lastUpdateTime: "2026-10-30T00:13:40Z"
      message: ReplicaSet "facebook-opt-125m-simulated-kserve-84b4ff656c" has successfully
        progressed.
      reason: NewReplicaSetAvailable
      status: "True"
      type: Progressing
    observedGeneration: 1
    readyReplicas: 1
    replicas: 1
    terminatingReplicas: 0
    updatedReplicas: 1
- apiVersion: apps/v1
  kind: Deployment
  metadata:
    annotations:
      deployment.kubernetes.io/revision: "1"
    creationTimestamp: "2026-10-30T00:13:30Z"
    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
    managedFields:
    - apiVersion: apps/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          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:ownerReferences:
            .: {}
            k:{"uid":"3ea4ebd3-3a0a-4482-99ff-4243a4e04f10"}: {}
        f:spec:
          f:progressDeadlineSeconds: {}
          f:replicas: {}
          f:revisionHistoryLimit: {}
          f:selector: {}
          f:strategy:
            f:rollingUpdate:
              .: {}
              f:maxSurge: {}
              f:maxUnavailable: {}
            f:type: {}
          f:template:
            f:metadata:
              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: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: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: manager
      operation: Update
      time: "2026-10-30T00:13:30Z"
    - apiVersion: apps/v1
      fieldsType: FieldsV1
      fieldsV1:
        f:metadata:
          f:annotations:
            .: {}
            f:deployment.kubernetes.io/revision: {}
        f:status:
          f:availableReplicas: {}
          f:conditions:
            .: {}
            k:{"type":"Available"}:
              .: {}
              f:lastTransitionTime: {}
              f:lastUpdateTime: {}
              f:message: {}
              f:reason: {}
              f:status: {}
              f:type: {}
            k:{"type":"Progressing"}:
              .: {}
              f:lastTransitionTime: {}
              f:lastUpdateTime: {}
              f:message: {}
              f:reason: {}
              f:status: {}
              f:type: {}
          f:observedGeneration: {}
          f:readyReplicas: {}
          f:replicas: {}
          f:terminatingReplicas: {}
          f:updatedReplicas: {}
      manager: kube-controller-manager
      operation: Update
      subresource: status
      time: "2026-10-30T00:34:08Z"
    name: premium-simulated-simulated-premium-kserve
    namespace: llm
    ownerReferences:
    - apiVersion: serving.kserve.io/v1alpha2
      blockOwnerDeletion: true
      controller: true
      kind: LLMInferenceService
      name: premium-simulated-simulated-premium
      uid: 3ea4ebd3-3a0a-4482-99ff-4243a4e04f10
    resourceVersion: "86655"
    uid: 81f96edc-34ae-4e71-96d8-67625dbc960b
  spec:
    progressDeadlineSeconds: 600
    replicas: 1
    revisionHistoryLimit: 10
    selector:
      matchLabels:
        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
    strategy:
      rollingUpdate:
        maxSurge: 25%
        maxUnavailable: 25%
      type: RollingUpdate
    template:
      metadata:
        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
      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
            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
        dnsPolicy: ClusterFirst
        restartPolicy: Always
        schedulerName: default-scheduler
        securityContext: {}
        terminationGracePeriodSeconds: 60
        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
  status:
    availableReplicas: 1
    conditions:
    - lastTransitionTime: "2026-10-30T00:13:30Z"
      lastUpdateTime: "2026-10-30T00:13:41Z"
      message: ReplicaSet "premium-simulated-simulated-premium-kserve-5f5c4b459f"
        has successfully progressed.
      reason: NewReplicaSetAvailable
      status: "True"
      type: Progressing
    - lastTransitionTime: "2026-10-30T00:34:08Z"
      lastUpdateTime: "2026-10-30T00:34:08Z"
      message: Deployment has minimum availability.
      reason: MinimumReplicasAvailable
      status: "True"
      type: Available
    observedGeneration: 1
    readyReplicas: 1
    replicas: 1
    terminatingReplicas: 0
    updatedReplicas: 1
kind: DeploymentList
metadata:
  resourceVersion: "111763"