---
apiVersion: serving.kserve.io/v1alpha2
kind: LLMInferenceServiceConfig
metadata:
  annotations:
    internal.config.kubernetes.io/previousKinds: LLMInferenceServiceConfig
    internal.config.kubernetes.io/previousNames: kserve-config-llm-scheduler
    internal.config.kubernetes.io/previousNamespaces: opendatahub
    serving.kserve.io/well-known-config: "true"
  creationTimestamp: "2026-11-03T19:59:38Z"
  finalizers:
  - serving.kserve.io/llmisvcconfig-finalizer
  generation: 3
  labels:
    platform.opendatahub.io/part-of: kserve
  managedFields:
  - apiVersion: serving.kserve.io/v1alpha2
    fieldsType: FieldsV1
    fieldsV1:
      f:metadata:
        f:annotations:
          f:internal.config.kubernetes.io/previousKinds: {}
          f:internal.config.kubernetes.io/previousNames: {}
          f:internal.config.kubernetes.io/previousNamespaces: {}
          f:serving.kserve.io/well-known-config: {}
        f:labels:
          f:platform.opendatahub.io/part-of: {}
      f:spec:
        f:router:
          f:scheduler:
            f:annotations:
              f:app.kubernetes.io/version: {}
              f:certificates.kserve.io/expiration-v2: {}
            f:pool:
              f:spec:
                f:endpointPickerRef:
                  f:failureMode: {}
                  f:kind: {}
                  f:name: {}
                  f:port:
                    f:number: {}
                f:selector:
                  f:matchLabels:
                    f:app.kubernetes.io/name: {}
                    f:app.kubernetes.io/part-of: {}
                    f:kserve.io/component: {}
                f:targetPorts: {}
            f:template:
              f:containers:
                k:{"name":"main"}:
                  .: {}
                  f:command: {}
                  f:env:
                    k:{"name":"SSL_CERT_DIR"}:
                      .: {}
                      f:name: {}
                      f:value: {}
                  f:image: {}
                  f:imagePullPolicy: {}
                  f:lifecycle:
                    f:preStop:
                      f:exec:
                        f:command: {}
                  f:livenessProbe:
                    f:failureThreshold: {}
                    f:grpc:
                      f:port: {}
                      f:service: {}
                    f:initialDelaySeconds: {}
                    f:periodSeconds: {}
                    f:successThreshold: {}
                    f:timeoutSeconds: {}
                  f:name: {}
                  f:ports:
                    k:{"containerPort":5557,"protocol":"TCP"}:
                      .: {}
                      f:containerPort: {}
                      f:name: {}
                      f:protocol: {}
                    k:{"containerPort":9002,"protocol":"TCP"}:
                      .: {}
                      f:containerPort: {}
                      f:name: {}
                      f:protocol: {}
                    k:{"containerPort":9003,"protocol":"TCP"}:
                      .: {}
                      f:containerPort: {}
                      f:name: {}
                      f:protocol: {}
                    k:{"containerPort":9090,"protocol":"TCP"}:
                      .: {}
                      f:containerPort: {}
                      f:name: {}
                      f:protocol: {}
                  f:readinessProbe:
                    f:failureThreshold: {}
                    f:grpc:
                      f:port: {}
                      f:service: {}
                    f:initialDelaySeconds: {}
                    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:terminationMessagePath: {}
                  f:terminationMessagePolicy: {}
                  f:volumeMounts:
                    k:{"mountPath":"/var/run/kserve/tls"}:
                      .: {}
                      f:mountPath: {}
                      f:name: {}
                      f:readOnly: {}
              f:dnsPolicy: {}
              f:restartPolicy: {}
              f:terminationGracePeriodSeconds: {}
              f:volumes:
                k:{"name":"tls-certs"}:
                  .: {}
                  f:name: {}
                  f:secret:
                    f:secretName: {}
    manager: kserve-module-controller
    operation: Apply
    time: "2026-11-03T19:59:58Z"
  - apiVersion: serving.kserve.io/v1alpha2
    fieldsType: FieldsV1
    fieldsV1:
      f:metadata:
        f:finalizers:
          .: {}
          v:"serving.kserve.io/llmisvcconfig-finalizer": {}
      f:spec:
        f:model:
          .: {}
          f:uri: {}
    manager: manager
    operation: Update
    time: "2026-11-03T19:59:38Z"
  - apiVersion: serving.kserve.io/v1alpha2
    fieldsType: FieldsV1
    fieldsV1:
      f:status:
        .: {}
        f:conditions: {}
        f:observedGeneration: {}
        f:referencedBy: {}
    manager: manager
    operation: Update
    subresource: status
    time: "2026-11-03T20:09:40Z"
  name: v3-6-0-ea-2-kserve-config-llm-scheduler
  namespace: opendatahub
  resourceVersion: "35354"
  uid: a73da612-214e-4057-b040-c7a010df297e
spec:
  model:
    uri: ""
  router:
    scheduler:
      annotations:
        app.kubernetes.io/version: 0.10.0
        certificates.kserve.io/expiration-v2: "true"
      pool:
        spec:
          appProtocol: http
          endpointPickerRef:
            failureMode: FailOpen
            group: ""
            kind: Service
            name: '{{ ChildName .ObjectMeta.Name `-epp-service` }}'
            port:
              number: 9002
          selector:
            matchLabels:
              app.kubernetes.io/name: '{{ .ObjectMeta.Name }}'
              app.kubernetes.io/part-of: llminferenceservice
              kserve.io/component: workload
          targetPorts:
          - number: 8000
      template:
        containers:
        - command:
          - /app/epp
          - --pool-name
          - '{{ ChildName .ObjectMeta.Name `-inference-pool` }}'
          - --pool-namespace
          - '{{ .ObjectMeta.Namespace }}'
          - --zap-encoder
          - json
          - --grpc-port
          - "9002"
          - --grpc-health-port
          - "9003"
          - '{{ if .GlobalConfig.EnableTLS }}--enable-cert-reload=true{{- end }}'
          - '{{ if .GlobalConfig.EnableTLS }}--secure-serving=true{{else}}--secure-serving=false{{-
            end }}'
          - '{{ if .GlobalConfig.EnableTLS }}--model-server-metrics-scheme=https{{-
            end }}'
          - '{{ if .GlobalConfig.EnableTLS }}--cert-path=/var/run/kserve/tls{{- end
            }}'
          env:
          - name: SSL_CERT_DIR
            value: /var/run/kserve/tls:/var/run/secrets/kubernetes.io/serviceaccount:/etc/pki/tls/certs
          image: quay.io/opendatahub/odh-llm-d-router-endpoint-picker@sha256:5fe24a45d9c8039b6db33e8be331dab766610e6eb60f9d2bec1ecc1153ca8827
          imagePullPolicy: IfNotPresent
          lifecycle:
            preStop:
              exec:
                command:
                - /bin/sleep
                - "15"
          livenessProbe:
            failureThreshold: 3
            grpc:
              port: 9003
              service: liveness
            initialDelaySeconds: 5
            periodSeconds: 10
            successThreshold: 1
            timeoutSeconds: 1
          name: main
          ports:
          - containerPort: 9002
            name: grpc
            protocol: TCP
          - containerPort: 9003
            name: grpc-health
            protocol: TCP
          - containerPort: 9090
            name: metrics
            protocol: TCP
          - containerPort: 5557
            name: zmq
            protocol: TCP
          readinessProbe:
            failureThreshold: 3
            grpc:
              port: 9003
              service: readiness
            initialDelaySeconds: 30
            periodSeconds: 10
            successThreshold: 1
            timeoutSeconds: 1
          resources:
            limits:
              cpu: 6
              memory: 16Gi
            requests:
              cpu: 1
              memory: 2Gi
          securityContext:
            allowPrivilegeEscalation: false
            capabilities:
              drop:
              - ALL
            readOnlyRootFilesystem: true
            runAsNonRoot: true
            seccompProfile:
              type: RuntimeDefault
          terminationMessagePath: /dev/termination-log
          terminationMessagePolicy: FallbackToLogsOnError
          volumeMounts:
          - mountPath: /var/run/kserve/tls
            name: tls-certs
            readOnly: true
        dnsPolicy: ClusterFirst
        restartPolicy: Always
        terminationGracePeriodSeconds: 60
        volumes:
        - name: tls-certs
          secret:
            secretName: '{{ ChildName .ObjectMeta.Name `-kserve-self-signed-certs`
              }}'
status:
  conditions:
  - lastTransitionTime: "2026-11-03T20:09:40Z"
    message: referenced by 8 LLMInferenceService(s)
    reason: InUse
    severity: Info
    status: "True"
    type: ConfigInUse
  - lastTransitionTime: "2026-11-03T19:59:39Z"
    status: "True"
    type: Ready
  observedGeneration: 3
  referencedBy:
  - name: e2e-distinct-2-simulated
    namespace: llm
  - name: e2e-distinct-simulated
    namespace: llm
  - name: e2e-embedding-simulated
    namespace: llm
  - name: e2e-trlp-test-simulated
    namespace: llm
  - name: e2e-unconfigured-facebook-opt-125m-simulated
    namespace: llm
  - name: echo-simulator
    namespace: llm
  - name: facebook-opt-125m-simulated
    namespace: llm
  - name: premium-simulated-simulated-premium
    namespace: llm