---
apiVersion: apps/v1
items:
- apiVersion: apps/v1
  kind: Deployment
  metadata:
    annotations:
      deployment.kubernetes.io/revision: "1"
    creationTimestamp: "2026-11-03T13:17:11Z"
    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":"55296d15-ea16-4fa8-a1a4-769f5c42a8d9"}: {}
        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: "2026-11-03T13:17:11Z"
    - 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: "2026-11-03T13:17:22Z"
    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: 55296d15-ea16-4fa8-a1a4-769f5c42a8d9
    resourceVersion: "20483"
    uid: a57f6dcb-5b73-422a-852a-a26a6749f9b3
  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: "2026-11-03T13:17:22Z"
      lastUpdateTime: "2026-11-03T13:17:22Z"
      message: Deployment has minimum availability.
      reason: MinimumReplicasAvailable
      status: "True"
      type: Available
    - lastTransitionTime: "2026-11-03T13:17:11Z"
      lastUpdateTime: "2026-11-03T13:17:22Z"
      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: "2026-11-03T13:17:13Z"
    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":"8121a3c9-9b71-4768-8c62-3d9c41fc9e9b"}: {}
        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: "2026-11-03T13:17:13Z"
    - 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: "2026-11-03T13:17:24Z"
    name: e2e-distinct-simulated-kserve
    namespace: llm
    ownerReferences:
    - apiVersion: serving.kserve.io/v1alpha2
      blockOwnerDeletion: true
      controller: true
      kind: LLMInferenceService
      name: e2e-distinct-simulated
      uid: 8121a3c9-9b71-4768-8c62-3d9c41fc9e9b
    resourceVersion: "20631"
    uid: c28ec4ac-76ba-4858-925d-c3467010ebe6
  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: "2026-11-03T13:17:24Z"
      lastUpdateTime: "2026-11-03T13:17:24Z"
      message: Deployment has minimum availability.
      reason: MinimumReplicasAvailable
      status: "True"
      type: Available
    - lastTransitionTime: "2026-11-03T13:17:13Z"
      lastUpdateTime: "2026-11-03T13:17:24Z"
      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: "2026-11-03T13:17:14Z"
    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":"9570acca-9924-45a8-b120-127a2b024555"}: {}
        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: "2026-11-03T13:17:14Z"
    - 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: "2026-11-03T13:17:26Z"
    name: e2e-embedding-simulated-kserve
    namespace: llm
    ownerReferences:
    - apiVersion: serving.kserve.io/v1alpha2
      blockOwnerDeletion: true
      controller: true
      kind: LLMInferenceService
      name: e2e-embedding-simulated
      uid: 9570acca-9924-45a8-b120-127a2b024555
    resourceVersion: "20715"
    uid: 5200286d-e9ea-4208-b51a-1979d8dc7f5e
  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: "2026-11-03T13:17:26Z"
      lastUpdateTime: "2026-11-03T13:17:26Z"
      message: Deployment has minimum availability.
      reason: MinimumReplicasAvailable
      status: "True"
      type: Available
    - lastTransitionTime: "2026-11-03T13:17:15Z"
      lastUpdateTime: "2026-11-03T13:17:26Z"
      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: "2026-11-03T13:17:16Z"
    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":"a79bc979-ffcc-48ae-92dc-57302ca48d7a"}: {}
        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: "2026-11-03T13:17:16Z"
    - 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: "2026-11-03T13:17:27Z"
    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: a79bc979-ffcc-48ae-92dc-57302ca48d7a
    resourceVersion: "20736"
    uid: 9dcf95ca-83ba-4f49-b3db-8d2b386ccc83
  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: "2026-11-03T13:17:27Z"
      lastUpdateTime: "2026-11-03T13:17:27Z"
      message: Deployment has minimum availability.
      reason: MinimumReplicasAvailable
      status: "True"
      type: Available
    - lastTransitionTime: "2026-11-03T13:17:16Z"
      lastUpdateTime: "2026-11-03T13:17:27Z"
      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: "2026-11-03T13:17:18Z"
    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":"71eca7c0-aeb2-41ea-880d-1bb3bd2e2e5a"}: {}
        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: "2026-11-03T13:17:18Z"
    - 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: "2026-11-03T13:17:29Z"
    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: 71eca7c0-aeb2-41ea-880d-1bb3bd2e2e5a
    resourceVersion: "20779"
    uid: 37eacb15-3f8c-4ac6-912c-2f147d5d96e7
  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: "2026-11-03T13:17:29Z"
      lastUpdateTime: "2026-11-03T13:17:29Z"
      message: Deployment has minimum availability.
      reason: MinimumReplicasAvailable
      status: "True"
      type: Available
    - lastTransitionTime: "2026-11-03T13:17:18Z"
      lastUpdateTime: "2026-11-03T13:17:29Z"
      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: "2026-11-03T13:17:18Z"
    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":"4465d18d-d950-450b-b799-539bb15eb2c2"}: {}
        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: "2026-11-03T13:17:18Z"
    - 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: "2026-11-03T13:17:29Z"
    name: echo-simulator-kserve
    namespace: llm
    ownerReferences:
    - apiVersion: serving.kserve.io/v1alpha2
      blockOwnerDeletion: true
      controller: true
      kind: LLMInferenceService
      name: echo-simulator
      uid: 4465d18d-d950-450b-b799-539bb15eb2c2
    resourceVersion: "20785"
    uid: 98da68c6-1091-4e51-b50d-35c994ddd26b
  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: "2026-11-03T13:17:29Z"
      lastUpdateTime: "2026-11-03T13:17:29Z"
      message: Deployment has minimum availability.
      reason: MinimumReplicasAvailable
      status: "True"
      type: Available
    - lastTransitionTime: "2026-11-03T13:17:18Z"
      lastUpdateTime: "2026-11-03T13:17:29Z"
      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: "2026-11-03T13:17:21Z"
    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":"cc3c7891-ddce-4ed7-81bf-8d122fb9f80d"}: {}
        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: "2026-11-03T13:17:21Z"
    - 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: "2026-11-03T13:17:32Z"
    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: cc3c7891-ddce-4ed7-81bf-8d122fb9f80d
    resourceVersion: "20837"
    uid: e48fae65-6eb6-436f-9b42-7ef6538936da
  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: "2026-11-03T13:17:32Z"
      lastUpdateTime: "2026-11-03T13:17:32Z"
      message: Deployment has minimum availability.
      reason: MinimumReplicasAvailable
      status: "True"
      type: Available
    - lastTransitionTime: "2026-11-03T13:17:21Z"
      lastUpdateTime: "2026-11-03T13:17:32Z"
      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: "2026-11-03T13:17:23Z"
    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":"99bd39d0-1ec2-44d5-b1ef-2de29fdf5c09"}: {}
        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: "2026-11-03T13:17:23Z"
    - 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: "2026-11-03T13:21:38Z"
    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: 99bd39d0-1ec2-44d5-b1ef-2de29fdf5c09
    resourceVersion: "35193"
    uid: 6e33722e-f17b-4fe3-a551-a9288f1b4ed5
  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: "2026-11-03T13:17:23Z"
      lastUpdateTime: "2026-11-03T13:17:34Z"
      message: ReplicaSet "premium-simulated-simulated-premium-kserve-5f5c4b459f"
        has successfully progressed.
      reason: NewReplicaSetAvailable
      status: "True"
      type: Progressing
    - lastTransitionTime: "2026-11-03T13:21:38Z"
      lastUpdateTime: "2026-11-03T13:21:38Z"
      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: "156980"