--- apiVersion: apps/v1 items: - apiVersion: apps/v1 kind: Deployment metadata: annotations: deployment.kubernetes.io/revision: "1" creationTimestamp: "2026-07-02T21:39:28Z" generation: 1 labels: app.kubernetes.io/component: llminferenceservice-workload app.kubernetes.io/name: auth-disabled-test 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":"b7a9c3b5-5fc9-4163-b586-81246b5294c8"}: {} 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:command: {} f:env: .: {} k:{"name":"HF_HUB_CACHE"}: .: {} f:name: {} f:value: {} k:{"name":"HOME"}: .: {} f:name: {} f:value: {} k:{"name":"TORCHINDUCTOR_CACHE_DIR"}: .: {} f:name: {} f:value: {} k:{"name":"USER"}: .: {} f:name: {} f:value: {} k:{"name":"VLLM_CPU_KVCACHE_SPACE"}: .: {} f:name: {} f:value: {} k:{"name":"VLLM_ENABLE_V1_MULTIPROCESSING"}: .: {} f:name: {} f:value: {} 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: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":"/mnt/models"}: .: {} f:mountPath: {} f:name: {} f:readOnly: {} 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:initContainers: .: {} k:{"name":"storage-initializer"}: .: {} f:args: {} f:env: .: {} k:{"name":"AWS_ACCESS_KEY_ID"}: .: {} f:name: {} f:valueFrom: .: {} f:secretKeyRef: {} k:{"name":"AWS_CA_BUNDLE"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_CA_BUNDLE_CONFIGMAP"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_ENDPOINT_URL"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_SECRET_ACCESS_KEY"}: .: {} f:name: {} f:valueFrom: .: {} f:secretKeyRef: {} k:{"name":"HF_HUB_ENABLE_HF_TRANSFER"}: .: {} f:name: {} f:value: {} k:{"name":"HF_XET_HIGH_PERFORMANCE"}: .: {} f:name: {} f:value: {} k:{"name":"HF_XET_NUM_CONCURRENT_RANGE_GETS"}: .: {} f:name: {} f:value: {} k:{"name":"S3_ENDPOINT"}: .: {} f:name: {} f:value: {} k:{"name":"S3_USE_HTTPS"}: .: {} f:name: {} f:value: {} k:{"name":"S3_VERIFY_SSL"}: .: {} f:name: {} f:value: {} f:image: {} f:imagePullPolicy: {} f:name: {} f:resources: .: {} f:limits: .: {} f:cpu: {} f:memory: {} f:requests: .: {} f:cpu: {} f:memory: {} f:terminationMessagePath: {} f:terminationMessagePolicy: {} f:volumeMounts: .: {} k:{"mountPath":"/mnt/models"}: .: {} f:mountPath: {} f:name: {} 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":"kserve-provision-location"}: .: {} 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: "2026-07-02T21:39:28Z" - 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:updatedReplicas: {} manager: kube-controller-manager operation: Update subresource: status time: "2026-07-02T21:41:38Z" name: auth-disabled-test-kserve namespace: kserve-ci-e2e-test ownerReferences: - apiVersion: serving.kserve.io/v1alpha2 blockOwnerDeletion: true controller: true kind: LLMInferenceService name: auth-disabled-test uid: b7a9c3b5-5fc9-4163-b586-81246b5294c8 resourceVersion: "30848" uid: 5a7bfb0e-c7e2-45bb-910e-d39a4e47e542 spec: progressDeadlineSeconds: 600 replicas: 1 revisionHistoryLimit: 10 selector: matchLabels: app.kubernetes.io/component: llminferenceservice-workload app.kubernetes.io/name: auth-disabled-test 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: auth-disabled-test app.kubernetes.io/part-of: llminferenceservice kserve.io/component: workload llm-d.ai/role: both spec: containers: - command: - /bin/bash - -c - |- if [ -f /etc/profile.d/ibm-aiu-setup.sh ]; then source /etc/profile.d/ibm-aiu-setup.sh fi if [ "$KSERVE_INFER_ROCE" = "true" ]; then echo "Trying to infer RoCE configs ... " grep -H . /sys/class/infiniband/*/ports/*/gids/* 2>/dev/null grep -H . /sys/class/infiniband/*/ports/*/gid_attrs/types/* 2>/dev/null cat /proc/driver/nvidia/params KSERVE_INFER_IB_GID_INDEX_GREP=${KSERVE_INFER_IB_GID_INDEX_GREP:-"RoCE v2"} echo "[Infer RoCE] Discovering active HCAs ..." active_hcas=() # Loop through all mlx5 devices found in sysfs for hca_dir in /sys/class/infiniband/mlx5_*; do # Ensure it's a directory before proceeding if [ -d "$hca_dir" ]; then hca_name=$(basename "$hca_dir") port_state_file="$hca_dir/ports/1/state" # Assume port 1 type_file="$hca_dir/ports/1/gid_attrs/types/*" echo "[Infer RoCE] Check if the port state file ${port_state_file} exists and contains 'ACTIVE'" if [ -f "$port_state_file" ] && grep -q "ACTIVE" "$port_state_file" && grep -q "${KSERVE_INFER_IB_GID_INDEX_GREP}" ${type_file} 2>/dev/null; then echo "[Infer RoCE] Found active HCA: $hca_name" active_hcas+=("$hca_name") else echo "[Infer RoCE] Skipping inactive or down HCA: $hca_name" fi fi done # Check if we found any active HCAs if [ ${#active_hcas[@]} -gt 0 ]; then # Join the array elements with a comma hca_port_pairs=() for hca in "${active_hcas[@]}"; do hca_port_pairs+=("${hca}:1") done active_hca_list=$(IFS=,; echo "${active_hcas[*]}") hca_port_pairs_list=$(IFS=,; echo "${hca_port_pairs[*]}") echo "[Infer RoCE] Setting active HCAs: ${active_hca_list}" export NCCL_IB_HCA=${NCCL_IB_HCA:-${active_hca_list}} export NVSHMEM_HCA_LIST=${NVSHMEM_HCA_LIST:-${hca_port_pairs_list}} export UCX_NET_DEVICES=${UCX_NET_DEVICES:-${hca_port_pairs_list}} echo "[Infer RoCE] NCCL_IB_HCA=${NCCL_IB_HCA}" echo "[Infer RoCE] NVSHMEM_HCA_LIST=${NVSHMEM_HCA_LIST}" echo "[Infer RoCE] UCX_NET_DEVICES=${UCX_NET_DEVICES}" else echo "[Infer RoCE] WARNING: No active RoCE HCAs found. NCCL_IB_HCA will not be set." fi if [ ${#active_hcas[@]} -gt 0 ]; then echo "[Infer RoCE] Finding GID_INDEX for each active HCA (SR-IOV compatible)..." # For SR-IOV environments, find the most common IPv4 RoCE v2 GID index across all HCAs declare -A gid_index_count declare -A hca_gid_index for hca_name in "${active_hcas[@]}"; do echo "[Infer RoCE] Processing HCA: ${hca_name}" # Find all RoCE v2 IPv4 GIDs for this HCA and count by index for tpath in /sys/class/infiniband/${hca_name}/ports/1/gid_attrs/types/*; do if grep -q "${KSERVE_INFER_IB_GID_INDEX_GREP}" "$tpath" 2>/dev/null; then idx=$(basename "$tpath") gid_file="/sys/class/infiniband/${hca_name}/ports/1/gids/${idx}" # Check for IPv4 GID (contains ffff:) if [ -f "$gid_file" ] && grep -q "ffff:" "$gid_file"; then gid_value=$(cat "$gid_file" 2>/dev/null || echo "") echo "[Infer RoCE] Found IPv4 RoCE v2 GID for ${hca_name}: index=${idx}, gid=${gid_value}" hca_gid_index["${hca_name}"]="${idx}" gid_index_count["${idx}"]=$((${gid_index_count["${idx}"]} + 1)) break # Use first found IPv4 GID per HCA fi fi done done # Find the most common GID index (most likely to be consistent across nodes) best_gid_index="" max_count=0 for idx in "${!gid_index_count[@]}"; do count=${gid_index_count["${idx}"]} echo "[Infer RoCE] GID_INDEX ${idx} found on ${count} HCAs" if [ $count -gt $max_count ]; then max_count=$count best_gid_index="$idx" fi done # Use deterministic fallback if tied - prefer index 3 (SR-IOV standard) if [ ${#gid_index_count[@]} -gt 1 ]; then echo "[Infer RoCE] Multiple GID indices found, selecting most common: ${best_gid_index}" # If there's a tie, prefer index 3 as it's most common in SR-IOV setups if [ -n "${gid_index_count['3']}" ] && [ "${gid_index_count['3']}" -eq "$max_count" ]; then best_gid_index="3" echo "[Infer RoCE] Using deterministic fallback: GID_INDEX=3 (SR-IOV standard)" fi fi # Check if GID_INDEX is already set via environment variables if [ -n "${NCCL_IB_GID_INDEX}" ]; then echo "[Infer RoCE] Using pre-configured NCCL_IB_GID_INDEX=${NCCL_IB_GID_INDEX} from environment" export NVSHMEM_IB_GID_INDEX=${NVSHMEM_IB_GID_INDEX:-$NCCL_IB_GID_INDEX} export UCX_IB_GID_INDEX=${UCX_IB_GID_INDEX:-$NCCL_IB_GID_INDEX} echo "[Infer RoCE] Using pre-configured GID_INDEX=${NCCL_IB_GID_INDEX} for NCCL, NVSHMEM, and UCX" elif [ -n "$best_gid_index" ]; then echo "[Infer RoCE] Selected GID_INDEX: ${best_gid_index} (found on ${max_count} HCAs)" export NCCL_IB_GID_INDEX=${NCCL_IB_GID_INDEX:-$best_gid_index} export NVSHMEM_IB_GID_INDEX=${NVSHMEM_IB_GID_INDEX:-$best_gid_index} export UCX_IB_GID_INDEX=${UCX_IB_GID_INDEX:-$best_gid_index} echo "[Infer RoCE] Exported GID_INDEX=${best_gid_index} for NCCL, NVSHMEM, and UCX" else echo "[Infer RoCE] ERROR: No valid IPv4 ${KSERVE_INFER_IB_GID_INDEX_GREP} GID_INDEX found on any HCA." fi else echo "[Infer RoCE] No active HCAs found, skipping GID_INDEX inference." fi fi # --disable-access-log-for-endpoints landed in vLLM 0.16.0 (vllm-project/vllm#30011). # Older versions still need the blanket --disable-uvicorn-access-log. ACCESS_LOG_ARGS="--disable-uvicorn-access-log" VLLM_VERSION=$(vllm --version 2>/dev/null | tail -1 | awk '{print $NF}') echo "[access-log-detect] vllm version='${VLLM_VERSION}'" if [[ "$VLLM_VERSION" =~ ^[0-9]+\.[0-9]+ ]] && [ "$(printf '%s\n%s\n' "0.16.0" "${VLLM_VERSION}" | sort -V | head -1)" = "0.16.0" ]; then ACCESS_LOG_ARGS="--disable-access-log-for-endpoints /health,/metrics,/ping" fi echo "[access-log-detect] selected ACCESS_LOG_ARGS='${ACCESS_LOG_ARGS}'" # --shutdown-timeout landed in vLLM 0.18.0 (vllm-project/vllm#36666). SHUTDOWN_TIMEOUT_ARGS="" if [[ "$VLLM_VERSION" =~ ^[0-9]+\.[0-9]+ ]] && [ "$(printf '%s\n%s\n' "0.18.0" "${VLLM_VERSION}" | sort -V | head -1)" = "0.18.0" ]; then SHUTDOWN_TIMEOUT_ARGS="--shutdown-timeout 40" fi eval "exec vllm serve /mnt/models \ --served-model-name "facebook/opt-125m" "publishers/kserve-ci-e2e-test/models/facebook/opt-125m" \ --port 8000 \ ${ACCESS_LOG_ARGS} \ ${SHUTDOWN_TIMEOUT_ARGS} \ --enable-ssl-refresh \ --ssl-certfile /var/run/kserve/tls/tls.crt \ --ssl-keyfile /var/run/kserve/tls/tls.key \ ${VLLM_ADDITIONAL_ARGS} \ $@" - -- env: - name: HOME value: /home - name: VLLM_LOGGING_LEVEL value: DEBUG - name: VLLM_CPU_KVCACHE_SPACE value: "1" - name: VLLM_ENABLE_V1_MULTIPROCESSING value: "0" - name: USER value: nonroot - name: TORCHINDUCTOR_CACHE_DIR value: /tmp/torchinductor-cache - name: HF_HUB_CACHE value: /models image: public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:v0.19.0 imagePullPolicy: IfNotPresent lifecycle: preStop: exec: command: - /bin/sleep - "15" livenessProbe: failureThreshold: 10 httpGet: path: /health port: 8000 scheme: HTTPS periodSeconds: 10 successThreshold: 1 timeoutSeconds: 1 name: main ports: - containerPort: 8000 protocol: TCP readinessProbe: failureThreshold: 2 httpGet: path: /health port: 8000 scheme: HTTPS periodSeconds: 1 successThreshold: 1 timeoutSeconds: 1 resources: limits: cpu: "2" memory: 7Gi requests: cpu: 200m memory: 2Gi securityContext: allowPrivilegeEscalation: false capabilities: drop: - ALL readOnlyRootFilesystem: true 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 - mountPath: /mnt/models name: kserve-provision-location readOnly: true dnsPolicy: ClusterFirst initContainers: - args: - hf://facebook/opt-125m - /mnt/models env: - name: AWS_ACCESS_KEY_ID valueFrom: secretKeyRef: key: AWS_ACCESS_KEY_ID name: seaweedfs-s3-creds - name: AWS_SECRET_ACCESS_KEY valueFrom: secretKeyRef: key: AWS_SECRET_ACCESS_KEY name: seaweedfs-s3-creds - name: S3_USE_HTTPS value: "0" - name: S3_ENDPOINT value: s3-service.kserve:8333 - name: AWS_ENDPOINT_URL value: http://s3-service.kserve:8333 - name: S3_VERIFY_SSL value: "0" - name: AWS_CA_BUNDLE value: /etc/ssl/custom-certs/cabundle.crt - name: AWS_CA_BUNDLE_CONFIGMAP value: odh-kserve-custom-ca-bundle - name: HF_HUB_ENABLE_HF_TRANSFER value: "1" - name: HF_XET_HIGH_PERFORMANCE value: "1" - name: HF_XET_NUM_CONCURRENT_RANGE_GETS value: "8" image: quay.io/opendatahub/kserve-storage-initializer@sha256:33f82c067d520fe749f73749cf76b5935fe69bd9eb141504c6660698c7947743 imagePullPolicy: IfNotPresent name: storage-initializer resources: limits: cpu: "1" memory: 24Gi requests: cpu: 100m memory: 100Mi terminationMessagePath: /dev/termination-log terminationMessagePolicy: FallbackToLogsOnError volumeMounts: - mountPath: /mnt/models name: kserve-provision-location 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: auth-disabled-test-kserve-self-signed-certs - emptyDir: {} name: kserve-provision-location status: availableReplicas: 1 conditions: - lastTransitionTime: "2026-07-02T21:41:38Z" lastUpdateTime: "2026-07-02T21:41:38Z" message: Deployment has minimum availability. reason: MinimumReplicasAvailable status: "True" type: Available - lastTransitionTime: "2026-07-02T21:39:28Z" lastUpdateTime: "2026-07-02T21:41:38Z" message: ReplicaSet "auth-disabled-test-kserve-5f858fd8bc" has successfully progressed. reason: NewReplicaSetAvailable status: "True" type: Progressing observedGeneration: 1 readyReplicas: 1 replicas: 1 updatedReplicas: 1 - apiVersion: apps/v1 kind: Deployment metadata: annotations: deployment.kubernetes.io/revision: "1" creationTimestamp: "2026-07-02T21:39:28Z" generation: 1 labels: app.kubernetes.io/component: llminferenceservice-router-scheduler app.kubernetes.io/name: auth-disabled-test app.kubernetes.io/part-of: llminferenceservice 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:ownerReferences: .: {} k:{"uid":"b7a9c3b5-5fc9-4163-b586-81246b5294c8"}: {} f:spec: f:progressDeadlineSeconds: {} f:replicas: {} f:revisionHistoryLimit: {} f:selector: {} f:strategy: f:type: {} f:template: f:metadata: f:annotations: .: {} f:app.kubernetes.io/version: {} f:certificates.kserve.io/expiration-v2: {} f:labels: .: {} f:app.kubernetes.io/component: {} f:app.kubernetes.io/name: {} f:app.kubernetes.io/part-of: {} f:spec: f:containers: k:{"name":"main"}: .: {} f:args: {} 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: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":"/tmp/tokenizer"}: .: {} f:mountPath: {} f:name: {} k:{"mountPath":"/var/run/kserve/tls"}: .: {} f:mountPath: {} f:name: {} f:readOnly: {} f:dnsPolicy: {} f:initContainers: .: {} k:{"name":"storage-initializer"}: .: {} f:args: {} f:env: .: {} k:{"name":"AWS_ACCESS_KEY_ID"}: .: {} f:name: {} f:valueFrom: .: {} f:secretKeyRef: {} k:{"name":"AWS_CA_BUNDLE"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_CA_BUNDLE_CONFIGMAP"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_ENDPOINT_URL"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_SECRET_ACCESS_KEY"}: .: {} f:name: {} f:valueFrom: .: {} f:secretKeyRef: {} k:{"name":"HF_HUB_ENABLE_HF_TRANSFER"}: .: {} f:name: {} f:value: {} k:{"name":"HF_XET_HIGH_PERFORMANCE"}: .: {} f:name: {} f:value: {} k:{"name":"HF_XET_NUM_CONCURRENT_RANGE_GETS"}: .: {} f:name: {} f:value: {} k:{"name":"S3_ENDPOINT"}: .: {} f:name: {} f:value: {} k:{"name":"S3_USE_HTTPS"}: .: {} f:name: {} f:value: {} k:{"name":"S3_VERIFY_SSL"}: .: {} f:name: {} f:value: {} k:{"name":"STORAGE_ALLOW_PATTERNS"}: .: {} f:name: {} f:value: {} f:image: {} f:imagePullPolicy: {} f:name: {} f:resources: .: {} f:limits: .: {} f:cpu: {} f:memory: {} f:requests: .: {} f:cpu: {} f:memory: {} f:terminationMessagePath: {} f:terminationMessagePolicy: {} f:volumeMounts: .: {} k:{"mountPath":"/mnt/models"}: .: {} f:mountPath: {} f:name: {} f:restartPolicy: {} f:schedulerName: {} f:securityContext: {} f:serviceAccount: {} f:serviceAccountName: {} f:terminationGracePeriodSeconds: {} f:volumes: .: {} k:{"name":"kserve-provision-location"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"tls-certs"}: .: {} f:name: {} f:secret: .: {} f:defaultMode: {} f:secretName: {} k:{"name":"tokenizer-cache"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"tokenizer-tmp"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"tokenizer-uds"}: .: {} f:emptyDir: {} f:name: {} manager: manager operation: Update time: "2026-07-02T21:39:28Z" - 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:updatedReplicas: {} manager: kube-controller-manager operation: Update subresource: status time: "2026-07-02T21:40:02Z" name: auth-disabled-test-kserve-router-scheduler namespace: kserve-ci-e2e-test ownerReferences: - apiVersion: serving.kserve.io/v1alpha2 blockOwnerDeletion: true controller: true kind: LLMInferenceService name: auth-disabled-test uid: b7a9c3b5-5fc9-4163-b586-81246b5294c8 resourceVersion: "29282" uid: 7c2b6e85-840d-4961-b4a8-b1f3968edeee spec: progressDeadlineSeconds: 600 replicas: 1 revisionHistoryLimit: 10 selector: matchLabels: app.kubernetes.io/component: llminferenceservice-router-scheduler app.kubernetes.io/name: auth-disabled-test app.kubernetes.io/part-of: llminferenceservice strategy: type: Recreate template: metadata: annotations: app.kubernetes.io/version: 0.9.0 certificates.kserve.io/expiration-v2: "true" labels: app.kubernetes.io/component: llminferenceservice-router-scheduler app.kubernetes.io/name: auth-disabled-test app.kubernetes.io/part-of: llminferenceservice spec: containers: - args: - --config-text - | apiVersion: inference.networking.x-k8s.io/v1alpha1 kind: EndpointPickerConfig plugins: - type: single-profile-handler - type: queue-scorer - type: prefix-cache-scorer - type: max-score-picker schedulingProfiles: - name: default plugins: - pluginRef: queue-scorer weight: 2 - pluginRef: prefix-cache-scorer weight: 3 - pluginRef: max-score-picker command: - /app/epp - --pool-name - auth-disabled-test-inference-pool - --pool-namespace - kserve-ci-e2e-test - --zap-encoder - json - --grpc-port - "9002" - --grpc-health-port - "9003" - --enable-cert-reload=true - --secure-serving=true - --model-server-metrics-scheme=https - --cert-path=/var/run/kserve/tls env: - name: SSL_CERT_DIR value: /var/run/kserve/tls:/var/run/secrets/kubernetes.io/serviceaccount:/etc/pki/tls/certs image: ghcr.io/llm-d/llm-d-router-endpoint-picker:v0.9.0-rc.2 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: requests: cpu: 256m memory: 500Mi 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 - mountPath: /tmp/tokenizer name: tokenizer-uds dnsPolicy: ClusterFirst initContainers: - args: - hf://facebook/opt-125m - /mnt/models env: - name: AWS_ACCESS_KEY_ID valueFrom: secretKeyRef: key: AWS_ACCESS_KEY_ID name: seaweedfs-s3-creds - name: AWS_SECRET_ACCESS_KEY valueFrom: secretKeyRef: key: AWS_SECRET_ACCESS_KEY name: seaweedfs-s3-creds - name: S3_USE_HTTPS value: "0" - name: S3_ENDPOINT value: s3-service.kserve:8333 - name: AWS_ENDPOINT_URL value: http://s3-service.kserve:8333 - name: S3_VERIFY_SSL value: "0" - name: AWS_CA_BUNDLE value: /etc/ssl/custom-certs/cabundle.crt - name: AWS_CA_BUNDLE_CONFIGMAP value: odh-kserve-custom-ca-bundle - name: HF_HUB_ENABLE_HF_TRANSFER value: "1" - name: HF_XET_HIGH_PERFORMANCE value: "1" - name: HF_XET_NUM_CONCURRENT_RANGE_GETS value: "8" - name: STORAGE_ALLOW_PATTERNS value: '["tokenizer.json", "tokenizer_config.json", "special_tokens_map.json", "vocab.json", "merges.txt", "config.json", "generation_config.json"]' image: quay.io/opendatahub/kserve-storage-initializer@sha256:33f82c067d520fe749f73749cf76b5935fe69bd9eb141504c6660698c7947743 imagePullPolicy: IfNotPresent name: storage-initializer resources: limits: cpu: "1" memory: 24Gi requests: cpu: 100m memory: 100Mi terminationMessagePath: /dev/termination-log terminationMessagePolicy: FallbackToLogsOnError volumeMounts: - mountPath: /mnt/models name: kserve-provision-location restartPolicy: Always schedulerName: default-scheduler securityContext: {} serviceAccount: auth-disabled-test-epp-sa serviceAccountName: auth-disabled-test-epp-sa terminationGracePeriodSeconds: 60 volumes: - name: tls-certs secret: defaultMode: 420 secretName: auth-disabled-test-kserve-self-signed-certs - emptyDir: {} name: tokenizer-uds - emptyDir: {} name: tokenizer-tmp - emptyDir: {} name: tokenizer-cache - emptyDir: {} name: kserve-provision-location status: availableReplicas: 1 conditions: - lastTransitionTime: "2026-07-02T21:40:02Z" lastUpdateTime: "2026-07-02T21:40:02Z" message: Deployment has minimum availability. reason: MinimumReplicasAvailable status: "True" type: Available - lastTransitionTime: "2026-07-02T21:39:28Z" lastUpdateTime: "2026-07-02T21:40:02Z" message: ReplicaSet "auth-disabled-test-kserve-router-scheduler-7dcc8bd5f8" has successfully progressed. reason: NewReplicaSetAvailable status: "True" type: Progressing observedGeneration: 1 readyReplicas: 1 replicas: 1 updatedReplicas: 1 - apiVersion: apps/v1 kind: Deployment metadata: annotations: deployment.kubernetes.io/revision: "1" creationTimestamp: "2026-07-02T22:10:51Z" generation: 1 labels: app.kubernetes.io/component: llminferenceservice-workload app.kubernetes.io/name: llmisvc-model-fb-opt-125m-route-50bc673d app.kubernetes.io/part-of: llminferenceservice kserve.io/component: workload llm-d.ai/role: decode 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":"7e453f7e-1de8-49d5-a8fa-024a9229a616"}: {} 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:command: {} f:env: .: {} k:{"name":"HF_HUB_CACHE"}: .: {} f:name: {} f:value: {} k:{"name":"HOME"}: .: {} f:name: {} f:value: {} k:{"name":"TORCHINDUCTOR_CACHE_DIR"}: .: {} f:name: {} f:value: {} k:{"name":"USER"}: .: {} f:name: {} f:value: {} k:{"name":"VLLM_CPU_KVCACHE_SPACE"}: .: {} f:name: {} f:value: {} k:{"name":"VLLM_ENABLE_V1_MULTIPROCESSING"}: .: {} f:name: {} f:value: {} 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:initialDelaySeconds: {} f:periodSeconds: {} f:successThreshold: {} f:timeoutSeconds: {} f:name: {} f:ports: .: {} k:{"containerPort":8001,"protocol":"TCP"}: .: {} f:containerPort: {} f:protocol: {} f:readinessProbe: .: {} f:failureThreshold: {} f:httpGet: .: {} f:path: {} f:port: {} f:scheme: {} 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: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":"/mnt/models"}: .: {} f:mountPath: {} f:name: {} f:readOnly: {} 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:initContainers: .: {} k:{"name":"llm-d-routing-sidecar"}: .: {} f:command: {} f:env: .: {} k:{"name":"INFERENCE_POOL_NAME"}: .: {} f:name: {} f:value: {} k:{"name":"INFERENCE_POOL_NAMESPACE"}: .: {} f:name: {} f:valueFrom: .: {} f:fieldRef: {} k:{"name":"SSL_CERT_DIR"}: .: {} f:name: {} f:value: {} f:image: {} f:imagePullPolicy: {} f:livenessProbe: .: {} f:failureThreshold: {} f:httpGet: .: {} f:path: {} f:port: {} f:scheme: {} f:initialDelaySeconds: {} f:periodSeconds: {} f:successThreshold: {} f:timeoutSeconds: {} f:name: {} f:ports: .: {} k:{"containerPort":8000,"protocol":"TCP"}: .: {} f:containerPort: {} f:protocol: {} f:readinessProbe: .: {} f:failureThreshold: {} f:httpGet: .: {} f:path: {} f:port: {} f:scheme: {} f:initialDelaySeconds: {} f:periodSeconds: {} f:successThreshold: {} f:timeoutSeconds: {} f:resources: {} f:restartPolicy: {} f:securityContext: .: {} f:allowPrivilegeEscalation: {} f:capabilities: .: {} f:drop: {} f:readOnlyRootFilesystem: {} f:runAsNonRoot: {} f:terminationMessagePath: {} f:terminationMessagePolicy: {} f:volumeMounts: .: {} k:{"mountPath":"/var/run/kserve/tls"}: .: {} f:mountPath: {} f:name: {} f:readOnly: {} k:{"name":"storage-initializer"}: .: {} f:args: {} f:env: .: {} k:{"name":"AWS_ACCESS_KEY_ID"}: .: {} f:name: {} f:valueFrom: .: {} f:secretKeyRef: {} k:{"name":"AWS_CA_BUNDLE"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_CA_BUNDLE_CONFIGMAP"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_ENDPOINT_URL"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_SECRET_ACCESS_KEY"}: .: {} f:name: {} f:valueFrom: .: {} f:secretKeyRef: {} k:{"name":"HF_HUB_ENABLE_HF_TRANSFER"}: .: {} f:name: {} f:value: {} k:{"name":"HF_XET_HIGH_PERFORMANCE"}: .: {} f:name: {} f:value: {} k:{"name":"HF_XET_NUM_CONCURRENT_RANGE_GETS"}: .: {} f:name: {} f:value: {} k:{"name":"S3_ENDPOINT"}: .: {} f:name: {} f:value: {} k:{"name":"S3_USE_HTTPS"}: .: {} f:name: {} f:value: {} k:{"name":"S3_VERIFY_SSL"}: .: {} f:name: {} f:value: {} f:image: {} f:imagePullPolicy: {} f:name: {} f:resources: .: {} f:limits: .: {} f:cpu: {} f:memory: {} f:requests: .: {} f:cpu: {} f:memory: {} f:terminationMessagePath: {} f:terminationMessagePolicy: {} f:volumeMounts: .: {} k:{"mountPath":"/mnt/models"}: .: {} f:mountPath: {} f:name: {} f:restartPolicy: {} f:schedulerName: {} f:securityContext: {} f:serviceAccount: {} f:serviceAccountName: {} f:terminationGracePeriodSeconds: {} f:volumes: .: {} k:{"name":"dshm"}: .: {} f:emptyDir: .: {} f:medium: {} f:sizeLimit: {} f:name: {} k:{"name":"home"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"kserve-provision-location"}: .: {} 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: "2026-07-02T22:10:51Z" - 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:updatedReplicas: {} manager: kube-controller-manager operation: Update subresource: status time: "2026-07-02T22:13:51Z" name: llmisvc-model-fb-opt-125m-route-50bc673d-kserve namespace: kserve-ci-e2e-test ownerReferences: - apiVersion: serving.kserve.io/v1alpha2 blockOwnerDeletion: true controller: true kind: LLMInferenceService name: llmisvc-model-fb-opt-125m-route-50bc673d uid: 7e453f7e-1de8-49d5-a8fa-024a9229a616 resourceVersion: "62697" uid: bf3cfc40-4bb8-4759-b3ba-9a362659c727 spec: progressDeadlineSeconds: 600 replicas: 1 revisionHistoryLimit: 10 selector: matchLabels: app.kubernetes.io/component: llminferenceservice-workload app.kubernetes.io/name: llmisvc-model-fb-opt-125m-route-50bc673d app.kubernetes.io/part-of: llminferenceservice kserve.io/component: workload llm-d.ai/role: decode strategy: rollingUpdate: maxSurge: 25% maxUnavailable: 25% type: RollingUpdate template: metadata: labels: app.kubernetes.io/component: llminferenceservice-workload app.kubernetes.io/name: llmisvc-model-fb-opt-125m-route-50bc673d app.kubernetes.io/part-of: llminferenceservice kserve.io/component: workload llm-d.ai/role: decode spec: containers: - command: - /bin/bash - -c - |- if [ -f /etc/profile.d/ibm-aiu-setup.sh ]; then source /etc/profile.d/ibm-aiu-setup.sh fi if [ "$KSERVE_INFER_ROCE" = "true" ]; then echo "Trying to infer RoCE configs ... " grep -H . /sys/class/infiniband/*/ports/*/gids/* 2>/dev/null grep -H . /sys/class/infiniband/*/ports/*/gid_attrs/types/* 2>/dev/null cat /proc/driver/nvidia/params KSERVE_INFER_IB_GID_INDEX_GREP=${KSERVE_INFER_IB_GID_INDEX_GREP:-"RoCE v2"} echo "[Infer RoCE] Discovering active HCAs ..." active_hcas=() # Loop through all mlx5 devices found in sysfs for hca_dir in /sys/class/infiniband/mlx5_*; do # Ensure it's a directory before proceeding if [ -d "$hca_dir" ]; then hca_name=$(basename "$hca_dir") port_state_file="$hca_dir/ports/1/state" # Assume port 1 type_file="$hca_dir/ports/1/gid_attrs/types/*" echo "[Infer RoCE] Check if the port state file ${port_state_file} exists and contains 'ACTIVE'" if [ -f "$port_state_file" ] && grep -q "ACTIVE" "$port_state_file" && grep -q "${KSERVE_INFER_IB_GID_INDEX_GREP}" ${type_file} 2>/dev/null; then echo "[Infer RoCE] Found active HCA: $hca_name" active_hcas+=("$hca_name") else echo "[Infer RoCE] Skipping inactive or down HCA: $hca_name" fi fi done # Check if we found any active HCAs if [ ${#active_hcas[@]} -gt 0 ]; then # Join the array elements with a comma hca_port_pairs=() for hca in "${active_hcas[@]}"; do hca_port_pairs+=("${hca}:1") done active_hca_list=$(IFS=,; echo "${active_hcas[*]}") hca_port_pairs_list=$(IFS=,; echo "${hca_port_pairs[*]}") echo "[Infer RoCE] Setting active HCAs: ${active_hca_list}" export NCCL_IB_HCA=${NCCL_IB_HCA:-${active_hca_list}} export NVSHMEM_HCA_LIST=${NVSHMEM_HCA_LIST:-${hca_port_pairs_list}} export UCX_NET_DEVICES=${UCX_NET_DEVICES:-${hca_port_pairs_list}} echo "[Infer RoCE] NCCL_IB_HCA=${NCCL_IB_HCA}" echo "[Infer RoCE] NVSHMEM_HCA_LIST=${NVSHMEM_HCA_LIST}" echo "[Infer RoCE] UCX_NET_DEVICES=${UCX_NET_DEVICES}" else echo "[Infer RoCE] WARNING: No active RoCE HCAs found. NCCL_IB_HCA will not be set." fi if [ ${#active_hcas[@]} -gt 0 ]; then echo "[Infer RoCE] Finding GID_INDEX for each active HCA (SR-IOV compatible)..." # For SR-IOV environments, find the most common IPv4 RoCE v2 GID index across all HCAs declare -A gid_index_count declare -A hca_gid_index for hca_name in "${active_hcas[@]}"; do echo "[Infer RoCE] Processing HCA: ${hca_name}" # Find all RoCE v2 IPv4 GIDs for this HCA and count by index for tpath in /sys/class/infiniband/${hca_name}/ports/1/gid_attrs/types/*; do if grep -q "${KSERVE_INFER_IB_GID_INDEX_GREP}" "$tpath" 2>/dev/null; then idx=$(basename "$tpath") gid_file="/sys/class/infiniband/${hca_name}/ports/1/gids/${idx}" # Check for IPv4 GID (contains ffff:) if [ -f "$gid_file" ] && grep -q "ffff:" "$gid_file"; then gid_value=$(cat "$gid_file" 2>/dev/null || echo "") echo "[Infer RoCE] Found IPv4 RoCE v2 GID for ${hca_name}: index=${idx}, gid=${gid_value}" hca_gid_index["${hca_name}"]="${idx}" gid_index_count["${idx}"]=$((${gid_index_count["${idx}"]} + 1)) break # Use first found IPv4 GID per HCA fi fi done done # Find the most common GID index (most likely to be consistent across nodes) best_gid_index="" max_count=0 for idx in "${!gid_index_count[@]}"; do count=${gid_index_count["${idx}"]} echo "[Infer RoCE] GID_INDEX ${idx} found on ${count} HCAs" if [ $count -gt $max_count ]; then max_count=$count best_gid_index="$idx" fi done # Use deterministic fallback if tied - prefer index 3 (SR-IOV standard) if [ ${#gid_index_count[@]} -gt 1 ]; then echo "[Infer RoCE] Multiple GID indices found, selecting most common: ${best_gid_index}" # If there's a tie, prefer index 3 as it's most common in SR-IOV setups if [ -n "${gid_index_count['3']}" ] && [ "${gid_index_count['3']}" -eq "$max_count" ]; then best_gid_index="3" echo "[Infer RoCE] Using deterministic fallback: GID_INDEX=3 (SR-IOV standard)" fi fi # Check if GID_INDEX is already set via environment variables if [ -n "${NCCL_IB_GID_INDEX}" ]; then echo "[Infer RoCE] Using pre-configured NCCL_IB_GID_INDEX=${NCCL_IB_GID_INDEX} from environment" export NVSHMEM_IB_GID_INDEX=${NVSHMEM_IB_GID_INDEX:-$NCCL_IB_GID_INDEX} export UCX_IB_GID_INDEX=${UCX_IB_GID_INDEX:-$NCCL_IB_GID_INDEX} echo "[Infer RoCE] Using pre-configured GID_INDEX=${NCCL_IB_GID_INDEX} for NCCL, NVSHMEM, and UCX" elif [ -n "$best_gid_index" ]; then echo "[Infer RoCE] Selected GID_INDEX: ${best_gid_index} (found on ${max_count} HCAs)" export NCCL_IB_GID_INDEX=${NCCL_IB_GID_INDEX:-$best_gid_index} export NVSHMEM_IB_GID_INDEX=${NVSHMEM_IB_GID_INDEX:-$best_gid_index} export UCX_IB_GID_INDEX=${UCX_IB_GID_INDEX:-$best_gid_index} echo "[Infer RoCE] Exported GID_INDEX=${best_gid_index} for NCCL, NVSHMEM, and UCX" else echo "[Infer RoCE] ERROR: No valid IPv4 ${KSERVE_INFER_IB_GID_INDEX_GREP} GID_INDEX found on any HCA." fi else echo "[Infer RoCE] No active HCAs found, skipping GID_INDEX inference." fi fi # --disable-access-log-for-endpoints landed in vLLM 0.16.0 (vllm-project/vllm#30011). # Older versions still need the blanket --disable-uvicorn-access-log. ACCESS_LOG_ARGS="--disable-uvicorn-access-log" VLLM_VERSION=$(vllm --version 2>/dev/null | tail -1 | awk '{print $NF}') echo "[access-log-detect] vllm version='${VLLM_VERSION}'" if [[ "$VLLM_VERSION" =~ ^[0-9]+\.[0-9]+ ]] && [ "$(printf '%s\n%s\n' "0.16.0" "${VLLM_VERSION}" | sort -V | head -1)" = "0.16.0" ]; then ACCESS_LOG_ARGS="--disable-access-log-for-endpoints /health,/metrics,/ping" fi echo "[access-log-detect] selected ACCESS_LOG_ARGS='${ACCESS_LOG_ARGS}'" # --shutdown-timeout landed in vLLM 0.18.0 (vllm-project/vllm#36666). SHUTDOWN_TIMEOUT_ARGS="" if [[ "$VLLM_VERSION" =~ ^[0-9]+\.[0-9]+ ]] && [ "$(printf '%s\n%s\n' "0.18.0" "${VLLM_VERSION}" | sort -V | head -1)" = "0.18.0" ]; then SHUTDOWN_TIMEOUT_ARGS="--shutdown-timeout 40" fi eval "exec vllm serve /mnt/models \ --served-model-name "facebook/opt-125m" "publishers/kserve-ci-e2e-test/models/facebook/opt-125m" \ --port 8001 \ ${ACCESS_LOG_ARGS} \ ${SHUTDOWN_TIMEOUT_ARGS} \ --enable-ssl-refresh \ --ssl-certfile /var/run/kserve/tls/tls.crt \ --ssl-keyfile /var/run/kserve/tls/tls.key \ ${VLLM_ADDITIONAL_ARGS} \ $@" - -- env: - name: HOME value: /home - name: VLLM_LOGGING_LEVEL value: DEBUG - name: VLLM_CPU_KVCACHE_SPACE value: "1" - name: VLLM_ENABLE_V1_MULTIPROCESSING value: "0" - name: USER value: nonroot - name: TORCHINDUCTOR_CACHE_DIR value: /tmp/torchinductor-cache - name: HF_HUB_CACHE value: /models image: public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:v0.19.0 imagePullPolicy: IfNotPresent lifecycle: preStop: exec: command: - /bin/sleep - "15" livenessProbe: failureThreshold: 8 httpGet: path: /health port: 8000 scheme: HTTPS initialDelaySeconds: 180 periodSeconds: 30 successThreshold: 1 timeoutSeconds: 30 name: main ports: - containerPort: 8001 protocol: TCP readinessProbe: failureThreshold: 3 httpGet: path: /health port: 8000 scheme: HTTPS initialDelaySeconds: 30 periodSeconds: 10 successThreshold: 1 timeoutSeconds: 5 resources: limits: cpu: "2" memory: 7Gi requests: cpu: 200m memory: 2Gi securityContext: allowPrivilegeEscalation: false capabilities: drop: - ALL readOnlyRootFilesystem: true runAsNonRoot: true seccompProfile: type: RuntimeDefault startupProbe: failureThreshold: 60 httpGet: path: /health port: 8001 scheme: HTTPS periodSeconds: 10 successThreshold: 1 timeoutSeconds: 1 terminationMessagePath: /dev/termination-log terminationMessagePolicy: FallbackToLogsOnError volumeMounts: - mountPath: /home name: home - mountPath: /tmp name: tmp-dir - mountPath: /dev/shm name: dshm - mountPath: /models name: model-cache - mountPath: /var/run/kserve/tls name: tls-certs readOnly: true - mountPath: /mnt/models name: kserve-provision-location readOnly: true dnsPolicy: ClusterFirst initContainers: - command: - /app/pd-sidecar - --port=8000 - --vllm-port=8001 - --kv-connector=nixlv2 - --enable-ssrf-protection=true - --pool-group=inference.networking.x-k8s.io - --inference-pool=kserve-ci-e2e-test/llmisvc-model-fb-opt-125m-route-50bc673d-inference-pool - --secure-proxy=true - --cert-path=/var/run/kserve/tls - --decoder-use-tls=true - --prefiller-use-tls=true env: - name: INFERENCE_POOL_NAMESPACE valueFrom: fieldRef: apiVersion: v1 fieldPath: metadata.namespace - name: SSL_CERT_DIR value: /var/run/kserve/tls:/var/run/secrets/kubernetes.io/serviceaccount:/etc/pki/tls/certs - name: INFERENCE_POOL_NAME value: llmisvc-model-fb-opt-125m-route-50bc673d-inference-pool image: ghcr.io/llm-d/llm-d-router-disagg-sidecar:v0.9.0-rc.2 imagePullPolicy: IfNotPresent livenessProbe: failureThreshold: 3 httpGet: path: /health port: 8000 scheme: HTTPS initialDelaySeconds: 10 periodSeconds: 10 successThreshold: 1 timeoutSeconds: 10 name: llm-d-routing-sidecar ports: - containerPort: 8000 protocol: TCP readinessProbe: failureThreshold: 10 httpGet: path: /health port: 8000 scheme: HTTPS initialDelaySeconds: 10 periodSeconds: 10 successThreshold: 1 timeoutSeconds: 5 resources: {} restartPolicy: Always securityContext: allowPrivilegeEscalation: false capabilities: drop: - ALL readOnlyRootFilesystem: true runAsNonRoot: true terminationMessagePath: /dev/termination-log terminationMessagePolicy: FallbackToLogsOnError volumeMounts: - mountPath: /var/run/kserve/tls name: tls-certs readOnly: true - args: - hf://facebook/opt-125m - /mnt/models env: - name: AWS_ACCESS_KEY_ID valueFrom: secretKeyRef: key: AWS_ACCESS_KEY_ID name: seaweedfs-s3-creds - name: AWS_SECRET_ACCESS_KEY valueFrom: secretKeyRef: key: AWS_SECRET_ACCESS_KEY name: seaweedfs-s3-creds - name: S3_USE_HTTPS value: "0" - name: S3_ENDPOINT value: s3-service.kserve:8333 - name: AWS_ENDPOINT_URL value: http://s3-service.kserve:8333 - name: S3_VERIFY_SSL value: "0" - name: AWS_CA_BUNDLE value: /etc/ssl/custom-certs/cabundle.crt - name: AWS_CA_BUNDLE_CONFIGMAP value: odh-kserve-custom-ca-bundle - name: HF_HUB_ENABLE_HF_TRANSFER value: "1" - name: HF_XET_HIGH_PERFORMANCE value: "1" - name: HF_XET_NUM_CONCURRENT_RANGE_GETS value: "8" image: quay.io/opendatahub/kserve-storage-initializer@sha256:33f82c067d520fe749f73749cf76b5935fe69bd9eb141504c6660698c7947743 imagePullPolicy: IfNotPresent name: storage-initializer resources: limits: cpu: "1" memory: 24Gi requests: cpu: 100m memory: 100Mi terminationMessagePath: /dev/termination-log terminationMessagePolicy: FallbackToLogsOnError volumeMounts: - mountPath: /mnt/models name: kserve-provision-location restartPolicy: Always schedulerName: default-scheduler securityContext: {} serviceAccount: llmisvc-model-fb-opt-125m-route-50bc673d-kserve serviceAccountName: llmisvc-model-fb-opt-125m-route-50bc673d-kserve terminationGracePeriodSeconds: 60 volumes: - emptyDir: {} name: home - emptyDir: {} name: tmp-dir - emptyDir: medium: Memory sizeLimit: 1Gi name: dshm - emptyDir: {} name: model-cache - name: tls-certs secret: defaultMode: 420 secretName: llmisv44d181485fad85e662eb092f3749502f-kserve-self-signed-certs - emptyDir: {} name: kserve-provision-location status: availableReplicas: 1 conditions: - lastTransitionTime: "2026-07-02T22:13:51Z" lastUpdateTime: "2026-07-02T22:13:51Z" message: Deployment has minimum availability. reason: MinimumReplicasAvailable status: "True" type: Available - lastTransitionTime: "2026-07-02T22:10:51Z" lastUpdateTime: "2026-07-02T22:13:51Z" message: ReplicaSet "llmisvc-model-fb-opt-125m-route-50bc673d-kserve-76cbf745cc" has successfully progressed. reason: NewReplicaSetAvailable status: "True" type: Progressing observedGeneration: 1 readyReplicas: 1 replicas: 1 updatedReplicas: 1 - apiVersion: apps/v1 kind: Deployment metadata: annotations: deployment.kubernetes.io/revision: "1" creationTimestamp: "2026-07-02T22:10:51Z" generation: 1 labels: app.kubernetes.io/component: llminferenceservice-workload-prefill app.kubernetes.io/name: llmisvc-model-fb-opt-125m-route-50bc673d app.kubernetes.io/part-of: llminferenceservice kserve.io/component: workload llm-d.ai/role: prefill 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":"7e453f7e-1de8-49d5-a8fa-024a9229a616"}: {} 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:command: {} f:env: .: {} k:{"name":"HF_HUB_CACHE"}: .: {} f:name: {} f:value: {} k:{"name":"HOME"}: .: {} f:name: {} f:value: {} k:{"name":"TORCHINDUCTOR_CACHE_DIR"}: .: {} f:name: {} f:value: {} k:{"name":"USER"}: .: {} f:name: {} f:value: {} k:{"name":"VLLM_CPU_KVCACHE_SPACE"}: .: {} f:name: {} f:value: {} k:{"name":"VLLM_ENABLE_V1_MULTIPROCESSING"}: .: {} f:name: {} f:value: {} 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:initialDelaySeconds: {} f:periodSeconds: {} f:successThreshold: {} f:timeoutSeconds: {} f:name: {} f:ports: .: {} k:{"containerPort":8000,"protocol":"TCP"}: .: {} f:containerPort: {} f:protocol: {} f:readinessProbe: .: {} f:failureThreshold: {} f:httpGet: .: {} f:path: {} f:port: {} f:scheme: {} 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: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":"/mnt/models"}: .: {} f:mountPath: {} f:name: {} f:readOnly: {} 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:initContainers: .: {} k:{"name":"storage-initializer"}: .: {} f:args: {} f:env: .: {} k:{"name":"AWS_ACCESS_KEY_ID"}: .: {} f:name: {} f:valueFrom: .: {} f:secretKeyRef: {} k:{"name":"AWS_CA_BUNDLE"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_CA_BUNDLE_CONFIGMAP"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_ENDPOINT_URL"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_SECRET_ACCESS_KEY"}: .: {} f:name: {} f:valueFrom: .: {} f:secretKeyRef: {} k:{"name":"HF_HUB_ENABLE_HF_TRANSFER"}: .: {} f:name: {} f:value: {} k:{"name":"HF_XET_HIGH_PERFORMANCE"}: .: {} f:name: {} f:value: {} k:{"name":"HF_XET_NUM_CONCURRENT_RANGE_GETS"}: .: {} f:name: {} f:value: {} k:{"name":"S3_ENDPOINT"}: .: {} f:name: {} f:value: {} k:{"name":"S3_USE_HTTPS"}: .: {} f:name: {} f:value: {} k:{"name":"S3_VERIFY_SSL"}: .: {} f:name: {} f:value: {} f:image: {} f:imagePullPolicy: {} f:name: {} f:resources: .: {} f:limits: .: {} f:cpu: {} f:memory: {} f:requests: .: {} f:cpu: {} f:memory: {} f:terminationMessagePath: {} f:terminationMessagePolicy: {} f:volumeMounts: .: {} k:{"mountPath":"/mnt/models"}: .: {} f:mountPath: {} f:name: {} 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":"kserve-provision-location"}: .: {} 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: "2026-07-02T22:10:51Z" - 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:updatedReplicas: {} manager: kube-controller-manager operation: Update subresource: status time: "2026-07-02T22:13:51Z" name: llmisvc-model-fb-opt-125m-route-50bc673d-kserve-prefill namespace: kserve-ci-e2e-test ownerReferences: - apiVersion: serving.kserve.io/v1alpha2 blockOwnerDeletion: true controller: true kind: LLMInferenceService name: llmisvc-model-fb-opt-125m-route-50bc673d uid: 7e453f7e-1de8-49d5-a8fa-024a9229a616 resourceVersion: "62707" uid: 5f632f3b-9f1b-414a-8509-f0b1fa435d24 spec: progressDeadlineSeconds: 600 replicas: 1 revisionHistoryLimit: 10 selector: matchLabels: app.kubernetes.io/component: llminferenceservice-workload-prefill app.kubernetes.io/name: llmisvc-model-fb-opt-125m-route-50bc673d app.kubernetes.io/part-of: llminferenceservice kserve.io/component: workload llm-d.ai/role: prefill strategy: rollingUpdate: maxSurge: 25% maxUnavailable: 25% type: RollingUpdate template: metadata: labels: app.kubernetes.io/component: llminferenceservice-workload-prefill app.kubernetes.io/name: llmisvc-model-fb-opt-125m-route-50bc673d app.kubernetes.io/part-of: llminferenceservice kserve.io/component: workload llm-d.ai/role: prefill spec: containers: - command: - /bin/bash - -c - |- if [ -f /etc/profile.d/ibm-aiu-setup.sh ]; then source /etc/profile.d/ibm-aiu-setup.sh fi if [ "$KSERVE_INFER_ROCE" = "true" ]; then echo "Trying to infer RoCE configs ... " grep -H . /sys/class/infiniband/*/ports/*/gids/* 2>/dev/null grep -H . /sys/class/infiniband/*/ports/*/gid_attrs/types/* 2>/dev/null cat /proc/driver/nvidia/params KSERVE_INFER_IB_GID_INDEX_GREP=${KSERVE_INFER_IB_GID_INDEX_GREP:-"RoCE v2"} echo "[Infer RoCE] Discovering active HCAs ..." active_hcas=() # Loop through all mlx5 devices found in sysfs for hca_dir in /sys/class/infiniband/mlx5_*; do # Ensure it's a directory before proceeding if [ -d "$hca_dir" ]; then hca_name=$(basename "$hca_dir") port_state_file="$hca_dir/ports/1/state" # Assume port 1 type_file="$hca_dir/ports/1/gid_attrs/types/*" echo "[Infer RoCE] Check if the port state file ${port_state_file} exists and contains 'ACTIVE'" if [ -f "$port_state_file" ] && grep -q "ACTIVE" "$port_state_file" && grep -q "${KSERVE_INFER_IB_GID_INDEX_GREP}" ${type_file} 2>/dev/null; then echo "[Infer RoCE] Found active HCA: $hca_name" active_hcas+=("$hca_name") else echo "[Infer RoCE] Skipping inactive or down HCA: $hca_name" fi fi done # Check if we found any active HCAs if [ ${#active_hcas[@]} -gt 0 ]; then # Join the array elements with a comma hca_port_pairs=() for hca in "${active_hcas[@]}"; do hca_port_pairs+=("${hca}:1") done active_hca_list=$(IFS=,; echo "${active_hcas[*]}") hca_port_pairs_list=$(IFS=,; echo "${hca_port_pairs[*]}") echo "[Infer RoCE] Setting active HCAs: ${active_hca_list}" export NCCL_IB_HCA=${NCCL_IB_HCA:-${active_hca_list}} export NVSHMEM_HCA_LIST=${NVSHMEM_HCA_LIST:-${hca_port_pairs_list}} export UCX_NET_DEVICES=${UCX_NET_DEVICES:-${hca_port_pairs_list}} echo "[Infer RoCE] NCCL_IB_HCA=${NCCL_IB_HCA}" echo "[Infer RoCE] NVSHMEM_HCA_LIST=${NVSHMEM_HCA_LIST}" echo "[Infer RoCE] UCX_NET_DEVICES=${UCX_NET_DEVICES}" else echo "[Infer RoCE] WARNING: No active RoCE HCAs found. NCCL_IB_HCA will not be set." fi if [ ${#active_hcas[@]} -gt 0 ]; then echo "[Infer RoCE] Finding GID_INDEX for each active HCA (SR-IOV compatible)..." # For SR-IOV environments, find the most common IPv4 RoCE v2 GID index across all HCAs declare -A gid_index_count declare -A hca_gid_index for hca_name in "${active_hcas[@]}"; do echo "[Infer RoCE] Processing HCA: ${hca_name}" # Find all RoCE v2 IPv4 GIDs for this HCA and count by index for tpath in /sys/class/infiniband/${hca_name}/ports/1/gid_attrs/types/*; do if grep -q "${KSERVE_INFER_IB_GID_INDEX_GREP}" "$tpath" 2>/dev/null; then idx=$(basename "$tpath") gid_file="/sys/class/infiniband/${hca_name}/ports/1/gids/${idx}" # Check for IPv4 GID (contains ffff:) if [ -f "$gid_file" ] && grep -q "ffff:" "$gid_file"; then gid_value=$(cat "$gid_file" 2>/dev/null || echo "") echo "[Infer RoCE] Found IPv4 RoCE v2 GID for ${hca_name}: index=${idx}, gid=${gid_value}" hca_gid_index["${hca_name}"]="${idx}" gid_index_count["${idx}"]=$((${gid_index_count["${idx}"]} + 1)) break # Use first found IPv4 GID per HCA fi fi done done # Find the most common GID index (most likely to be consistent across nodes) best_gid_index="" max_count=0 for idx in "${!gid_index_count[@]}"; do count=${gid_index_count["${idx}"]} echo "[Infer RoCE] GID_INDEX ${idx} found on ${count} HCAs" if [ $count -gt $max_count ]; then max_count=$count best_gid_index="$idx" fi done # Use deterministic fallback if tied - prefer index 3 (SR-IOV standard) if [ ${#gid_index_count[@]} -gt 1 ]; then echo "[Infer RoCE] Multiple GID indices found, selecting most common: ${best_gid_index}" # If there's a tie, prefer index 3 as it's most common in SR-IOV setups if [ -n "${gid_index_count['3']}" ] && [ "${gid_index_count['3']}" -eq "$max_count" ]; then best_gid_index="3" echo "[Infer RoCE] Using deterministic fallback: GID_INDEX=3 (SR-IOV standard)" fi fi # Check if GID_INDEX is already set via environment variables if [ -n "${NCCL_IB_GID_INDEX}" ]; then echo "[Infer RoCE] Using pre-configured NCCL_IB_GID_INDEX=${NCCL_IB_GID_INDEX} from environment" export NVSHMEM_IB_GID_INDEX=${NVSHMEM_IB_GID_INDEX:-$NCCL_IB_GID_INDEX} export UCX_IB_GID_INDEX=${UCX_IB_GID_INDEX:-$NCCL_IB_GID_INDEX} echo "[Infer RoCE] Using pre-configured GID_INDEX=${NCCL_IB_GID_INDEX} for NCCL, NVSHMEM, and UCX" elif [ -n "$best_gid_index" ]; then echo "[Infer RoCE] Selected GID_INDEX: ${best_gid_index} (found on ${max_count} HCAs)" export NCCL_IB_GID_INDEX=${NCCL_IB_GID_INDEX:-$best_gid_index} export NVSHMEM_IB_GID_INDEX=${NVSHMEM_IB_GID_INDEX:-$best_gid_index} export UCX_IB_GID_INDEX=${UCX_IB_GID_INDEX:-$best_gid_index} echo "[Infer RoCE] Exported GID_INDEX=${best_gid_index} for NCCL, NVSHMEM, and UCX" else echo "[Infer RoCE] ERROR: No valid IPv4 ${KSERVE_INFER_IB_GID_INDEX_GREP} GID_INDEX found on any HCA." fi else echo "[Infer RoCE] No active HCAs found, skipping GID_INDEX inference." fi fi # --disable-access-log-for-endpoints landed in vLLM 0.16.0 (vllm-project/vllm#30011). # Older versions still need the blanket --disable-uvicorn-access-log. ACCESS_LOG_ARGS="--disable-uvicorn-access-log" VLLM_VERSION=$(vllm --version 2>/dev/null | tail -1 | awk '{print $NF}') echo "[access-log-detect] vllm version='${VLLM_VERSION}'" if [[ "$VLLM_VERSION" =~ ^[0-9]+\.[0-9]+ ]] && [ "$(printf '%s\n%s\n' "0.16.0" "${VLLM_VERSION}" | sort -V | head -1)" = "0.16.0" ]; then ACCESS_LOG_ARGS="--disable-access-log-for-endpoints /health,/metrics,/ping" fi echo "[access-log-detect] selected ACCESS_LOG_ARGS='${ACCESS_LOG_ARGS}'" # --shutdown-timeout landed in vLLM 0.18.0 (vllm-project/vllm#36666). SHUTDOWN_TIMEOUT_ARGS="" if [[ "$VLLM_VERSION" =~ ^[0-9]+\.[0-9]+ ]] && [ "$(printf '%s\n%s\n' "0.18.0" "${VLLM_VERSION}" | sort -V | head -1)" = "0.18.0" ]; then SHUTDOWN_TIMEOUT_ARGS="--shutdown-timeout 40" fi eval "exec vllm serve /mnt/models \ --served-model-name "facebook/opt-125m" \ --port 8000 \ ${ACCESS_LOG_ARGS} \ ${SHUTDOWN_TIMEOUT_ARGS} \ --enable-ssl-refresh \ --ssl-certfile /var/run/kserve/tls/tls.crt \ --ssl-keyfile /var/run/kserve/tls/tls.key \ ${VLLM_ADDITIONAL_ARGS} \ $@" - -- env: - name: HOME value: /home - name: VLLM_LOGGING_LEVEL value: DEBUG - name: VLLM_CPU_KVCACHE_SPACE value: "1" - name: VLLM_ENABLE_V1_MULTIPROCESSING value: "0" - name: USER value: nonroot - name: TORCHINDUCTOR_CACHE_DIR value: /tmp/torchinductor-cache - name: HF_HUB_CACHE value: /models image: public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:v0.19.0 imagePullPolicy: IfNotPresent lifecycle: preStop: exec: command: - /bin/sleep - "15" livenessProbe: failureThreshold: 8 httpGet: path: /health port: 8000 scheme: HTTPS initialDelaySeconds: 180 periodSeconds: 30 successThreshold: 1 timeoutSeconds: 30 name: main ports: - containerPort: 8000 protocol: TCP readinessProbe: failureThreshold: 3 httpGet: path: /health port: 8000 scheme: HTTPS initialDelaySeconds: 30 periodSeconds: 10 successThreshold: 1 timeoutSeconds: 5 resources: limits: cpu: "2" memory: 7Gi requests: cpu: 200m memory: 2Gi securityContext: allowPrivilegeEscalation: false capabilities: drop: - ALL readOnlyRootFilesystem: true 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: File volumeMounts: - mountPath: /home name: home - mountPath: /tmp name: tmp-dir - mountPath: /dev/shm name: dshm - mountPath: /models name: model-cache - mountPath: /var/run/kserve/tls name: tls-certs readOnly: true - mountPath: /mnt/models name: kserve-provision-location readOnly: true dnsPolicy: ClusterFirst initContainers: - args: - hf://facebook/opt-125m - /mnt/models env: - name: AWS_ACCESS_KEY_ID valueFrom: secretKeyRef: key: AWS_ACCESS_KEY_ID name: seaweedfs-s3-creds - name: AWS_SECRET_ACCESS_KEY valueFrom: secretKeyRef: key: AWS_SECRET_ACCESS_KEY name: seaweedfs-s3-creds - name: S3_USE_HTTPS value: "0" - name: S3_ENDPOINT value: s3-service.kserve:8333 - name: AWS_ENDPOINT_URL value: http://s3-service.kserve:8333 - name: S3_VERIFY_SSL value: "0" - name: AWS_CA_BUNDLE value: /etc/ssl/custom-certs/cabundle.crt - name: AWS_CA_BUNDLE_CONFIGMAP value: odh-kserve-custom-ca-bundle - name: HF_HUB_ENABLE_HF_TRANSFER value: "1" - name: HF_XET_HIGH_PERFORMANCE value: "1" - name: HF_XET_NUM_CONCURRENT_RANGE_GETS value: "8" image: quay.io/opendatahub/kserve-storage-initializer@sha256:33f82c067d520fe749f73749cf76b5935fe69bd9eb141504c6660698c7947743 imagePullPolicy: IfNotPresent name: storage-initializer resources: limits: cpu: "1" memory: 24Gi requests: cpu: 100m memory: 100Mi terminationMessagePath: /dev/termination-log terminationMessagePolicy: FallbackToLogsOnError volumeMounts: - mountPath: /mnt/models name: kserve-provision-location restartPolicy: Always schedulerName: default-scheduler securityContext: {} terminationGracePeriodSeconds: 60 volumes: - emptyDir: {} name: home - emptyDir: {} name: tmp-dir - emptyDir: medium: Memory sizeLimit: 1Gi name: dshm - emptyDir: {} name: model-cache - name: tls-certs secret: defaultMode: 420 secretName: llmisv44d181485fad85e662eb092f3749502f-kserve-self-signed-certs - emptyDir: {} name: kserve-provision-location status: availableReplicas: 1 conditions: - lastTransitionTime: "2026-07-02T22:13:51Z" lastUpdateTime: "2026-07-02T22:13:51Z" message: Deployment has minimum availability. reason: MinimumReplicasAvailable status: "True" type: Available - lastTransitionTime: "2026-07-02T22:10:51Z" lastUpdateTime: "2026-07-02T22:13:51Z" message: ReplicaSet "llmisvc-model-fb-opt-125m-route-50bc673d-kserve-prefill-649c8bb55d" has successfully progressed. reason: NewReplicaSetAvailable status: "True" type: Progressing observedGeneration: 1 readyReplicas: 1 replicas: 1 updatedReplicas: 1 - apiVersion: apps/v1 kind: Deployment metadata: annotations: deployment.kubernetes.io/revision: "1" creationTimestamp: "2026-07-02T21:41:10Z" generation: 1 labels: app.kubernetes.io/component: llminferenceservice-workload app.kubernetes.io/name: llmisvc-model-fb-opt-125m-with-7ca60146 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":"9e661501-7bd8-4942-8472-a92b8c944b91"}: {} 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":"TORCHINDUCTOR_CACHE_DIR"}: .: {} f:name: {} f:value: {} k:{"name":"USER"}: .: {} f:name: {} f:value: {} k:{"name":"VLLM_CPU_KVCACHE_SPACE"}: .: {} f:name: {} f:value: {} k:{"name":"VLLM_ENABLE_V1_MULTIPROCESSING"}: .: {} f:name: {} f:value: {} 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: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":"/mnt"}: .: {} f:mountPath: {} f:name: {} f:readOnly: {} 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:initContainers: .: {} k:{"name":"storage-initializer"}: .: {} f:args: {} f:env: .: {} k:{"name":"AWS_ACCESS_KEY_ID"}: .: {} f:name: {} f:valueFrom: .: {} f:secretKeyRef: {} k:{"name":"AWS_CA_BUNDLE"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_CA_BUNDLE_CONFIGMAP"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_ENDPOINT_URL"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_SECRET_ACCESS_KEY"}: .: {} f:name: {} f:valueFrom: .: {} f:secretKeyRef: {} k:{"name":"CA_BUNDLE_CONFIGMAP_NAME"}: .: {} f:name: {} f:value: {} k:{"name":"CA_BUNDLE_VOLUME_MOUNT_POINT"}: .: {} f:name: {} f:value: {} k:{"name":"HF_HUB_ENABLE_HF_TRANSFER"}: .: {} f:name: {} f:value: {} k:{"name":"HF_XET_HIGH_PERFORMANCE"}: .: {} f:name: {} f:value: {} k:{"name":"HF_XET_NUM_CONCURRENT_RANGE_GETS"}: .: {} f:name: {} f:value: {} k:{"name":"S3_ENDPOINT"}: .: {} f:name: {} f:value: {} k:{"name":"S3_USE_HTTPS"}: .: {} f:name: {} f:value: {} k:{"name":"S3_VERIFY_SSL"}: .: {} f:name: {} f:value: {} f:image: {} f:imagePullPolicy: {} f:name: {} f:resources: .: {} f:limits: .: {} f:cpu: {} f:memory: {} f:requests: .: {} f:cpu: {} f:memory: {} f:terminationMessagePath: {} f:terminationMessagePolicy: {} f:volumeMounts: .: {} k:{"mountPath":"/etc/ssl/custom-certs"}: .: {} f:mountPath: {} f:name: {} f:readOnly: {} k:{"mountPath":"/mnt"}: .: {} f:mountPath: {} f:name: {} f:restartPolicy: {} f:schedulerName: {} f:securityContext: {} f:terminationGracePeriodSeconds: {} f:volumes: .: {} k:{"name":"cabundle-cert"}: .: {} f:configMap: .: {} f:defaultMode: {} f:name: {} f:name: {} k:{"name":"dshm"}: .: {} f:emptyDir: .: {} f:medium: {} f:sizeLimit: {} f:name: {} k:{"name":"home"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"kserve-provision-location"}: .: {} 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: "2026-07-02T21:41:10Z" - 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:updatedReplicas: {} manager: kube-controller-manager operation: Update subresource: status time: "2026-07-02T21:42:41Z" name: llmisvc-model-fb-opt-125m-with-7ca60146-kserve namespace: kserve-ci-e2e-test ownerReferences: - apiVersion: serving.kserve.io/v1alpha2 blockOwnerDeletion: true controller: true kind: LLMInferenceService name: llmisvc-model-fb-opt-125m-with-7ca60146 uid: 9e661501-7bd8-4942-8472-a92b8c944b91 resourceVersion: "31960" uid: 95e7d282-bea8-4929-8d94-525663751b6d spec: progressDeadlineSeconds: 600 replicas: 1 revisionHistoryLimit: 10 selector: matchLabels: app.kubernetes.io/component: llminferenceservice-workload app.kubernetes.io/name: llmisvc-model-fb-opt-125m-with-7ca60146 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: llmisvc-model-fb-opt-125m-with-7ca60146 app.kubernetes.io/part-of: llminferenceservice kserve.io/component: workload llm-d.ai/role: both spec: containers: - args: - --enable-lora - --lora-modules - '''{"name":"lora-adapter-1","path":"/mnt/lora/lora-adapter-1"}''' - '''{"name":"publishers/kserve-ci-e2e-test/models/lora-adapter-1","path":"/mnt/lora/lora-adapter-1"}''' command: - /bin/bash - -c - |- if [ -f /etc/profile.d/ibm-aiu-setup.sh ]; then source /etc/profile.d/ibm-aiu-setup.sh fi if [ "$KSERVE_INFER_ROCE" = "true" ]; then echo "Trying to infer RoCE configs ... " grep -H . /sys/class/infiniband/*/ports/*/gids/* 2>/dev/null grep -H . /sys/class/infiniband/*/ports/*/gid_attrs/types/* 2>/dev/null cat /proc/driver/nvidia/params KSERVE_INFER_IB_GID_INDEX_GREP=${KSERVE_INFER_IB_GID_INDEX_GREP:-"RoCE v2"} echo "[Infer RoCE] Discovering active HCAs ..." active_hcas=() # Loop through all mlx5 devices found in sysfs for hca_dir in /sys/class/infiniband/mlx5_*; do # Ensure it's a directory before proceeding if [ -d "$hca_dir" ]; then hca_name=$(basename "$hca_dir") port_state_file="$hca_dir/ports/1/state" # Assume port 1 type_file="$hca_dir/ports/1/gid_attrs/types/*" echo "[Infer RoCE] Check if the port state file ${port_state_file} exists and contains 'ACTIVE'" if [ -f "$port_state_file" ] && grep -q "ACTIVE" "$port_state_file" && grep -q "${KSERVE_INFER_IB_GID_INDEX_GREP}" ${type_file} 2>/dev/null; then echo "[Infer RoCE] Found active HCA: $hca_name" active_hcas+=("$hca_name") else echo "[Infer RoCE] Skipping inactive or down HCA: $hca_name" fi fi done # Check if we found any active HCAs if [ ${#active_hcas[@]} -gt 0 ]; then # Join the array elements with a comma hca_port_pairs=() for hca in "${active_hcas[@]}"; do hca_port_pairs+=("${hca}:1") done active_hca_list=$(IFS=,; echo "${active_hcas[*]}") hca_port_pairs_list=$(IFS=,; echo "${hca_port_pairs[*]}") echo "[Infer RoCE] Setting active HCAs: ${active_hca_list}" export NCCL_IB_HCA=${NCCL_IB_HCA:-${active_hca_list}} export NVSHMEM_HCA_LIST=${NVSHMEM_HCA_LIST:-${hca_port_pairs_list}} export UCX_NET_DEVICES=${UCX_NET_DEVICES:-${hca_port_pairs_list}} echo "[Infer RoCE] NCCL_IB_HCA=${NCCL_IB_HCA}" echo "[Infer RoCE] NVSHMEM_HCA_LIST=${NVSHMEM_HCA_LIST}" echo "[Infer RoCE] UCX_NET_DEVICES=${UCX_NET_DEVICES}" else echo "[Infer RoCE] WARNING: No active RoCE HCAs found. NCCL_IB_HCA will not be set." fi if [ ${#active_hcas[@]} -gt 0 ]; then echo "[Infer RoCE] Finding GID_INDEX for each active HCA (SR-IOV compatible)..." # For SR-IOV environments, find the most common IPv4 RoCE v2 GID index across all HCAs declare -A gid_index_count declare -A hca_gid_index for hca_name in "${active_hcas[@]}"; do echo "[Infer RoCE] Processing HCA: ${hca_name}" # Find all RoCE v2 IPv4 GIDs for this HCA and count by index for tpath in /sys/class/infiniband/${hca_name}/ports/1/gid_attrs/types/*; do if grep -q "${KSERVE_INFER_IB_GID_INDEX_GREP}" "$tpath" 2>/dev/null; then idx=$(basename "$tpath") gid_file="/sys/class/infiniband/${hca_name}/ports/1/gids/${idx}" # Check for IPv4 GID (contains ffff:) if [ -f "$gid_file" ] && grep -q "ffff:" "$gid_file"; then gid_value=$(cat "$gid_file" 2>/dev/null || echo "") echo "[Infer RoCE] Found IPv4 RoCE v2 GID for ${hca_name}: index=${idx}, gid=${gid_value}" hca_gid_index["${hca_name}"]="${idx}" gid_index_count["${idx}"]=$((${gid_index_count["${idx}"]} + 1)) break # Use first found IPv4 GID per HCA fi fi done done # Find the most common GID index (most likely to be consistent across nodes) best_gid_index="" max_count=0 for idx in "${!gid_index_count[@]}"; do count=${gid_index_count["${idx}"]} echo "[Infer RoCE] GID_INDEX ${idx} found on ${count} HCAs" if [ $count -gt $max_count ]; then max_count=$count best_gid_index="$idx" fi done # Use deterministic fallback if tied - prefer index 3 (SR-IOV standard) if [ ${#gid_index_count[@]} -gt 1 ]; then echo "[Infer RoCE] Multiple GID indices found, selecting most common: ${best_gid_index}" # If there's a tie, prefer index 3 as it's most common in SR-IOV setups if [ -n "${gid_index_count['3']}" ] && [ "${gid_index_count['3']}" -eq "$max_count" ]; then best_gid_index="3" echo "[Infer RoCE] Using deterministic fallback: GID_INDEX=3 (SR-IOV standard)" fi fi # Check if GID_INDEX is already set via environment variables if [ -n "${NCCL_IB_GID_INDEX}" ]; then echo "[Infer RoCE] Using pre-configured NCCL_IB_GID_INDEX=${NCCL_IB_GID_INDEX} from environment" export NVSHMEM_IB_GID_INDEX=${NVSHMEM_IB_GID_INDEX:-$NCCL_IB_GID_INDEX} export UCX_IB_GID_INDEX=${UCX_IB_GID_INDEX:-$NCCL_IB_GID_INDEX} echo "[Infer RoCE] Using pre-configured GID_INDEX=${NCCL_IB_GID_INDEX} for NCCL, NVSHMEM, and UCX" elif [ -n "$best_gid_index" ]; then echo "[Infer RoCE] Selected GID_INDEX: ${best_gid_index} (found on ${max_count} HCAs)" export NCCL_IB_GID_INDEX=${NCCL_IB_GID_INDEX:-$best_gid_index} export NVSHMEM_IB_GID_INDEX=${NVSHMEM_IB_GID_INDEX:-$best_gid_index} export UCX_IB_GID_INDEX=${UCX_IB_GID_INDEX:-$best_gid_index} echo "[Infer RoCE] Exported GID_INDEX=${best_gid_index} for NCCL, NVSHMEM, and UCX" else echo "[Infer RoCE] ERROR: No valid IPv4 ${KSERVE_INFER_IB_GID_INDEX_GREP} GID_INDEX found on any HCA." fi else echo "[Infer RoCE] No active HCAs found, skipping GID_INDEX inference." fi fi # --disable-access-log-for-endpoints landed in vLLM 0.16.0 (vllm-project/vllm#30011). # Older versions still need the blanket --disable-uvicorn-access-log. ACCESS_LOG_ARGS="--disable-uvicorn-access-log" VLLM_VERSION=$(vllm --version 2>/dev/null | tail -1 | awk '{print $NF}') echo "[access-log-detect] vllm version='${VLLM_VERSION}'" if [[ "$VLLM_VERSION" =~ ^[0-9]+\.[0-9]+ ]] && [ "$(printf '%s\n%s\n' "0.16.0" "${VLLM_VERSION}" | sort -V | head -1)" = "0.16.0" ]; then ACCESS_LOG_ARGS="--disable-access-log-for-endpoints /health,/metrics,/ping" fi echo "[access-log-detect] selected ACCESS_LOG_ARGS='${ACCESS_LOG_ARGS}'" # --shutdown-timeout landed in vLLM 0.18.0 (vllm-project/vllm#36666). SHUTDOWN_TIMEOUT_ARGS="" if [[ "$VLLM_VERSION" =~ ^[0-9]+\.[0-9]+ ]] && [ "$(printf '%s\n%s\n' "0.18.0" "${VLLM_VERSION}" | sort -V | head -1)" = "0.18.0" ]; then SHUTDOWN_TIMEOUT_ARGS="--shutdown-timeout 40" fi eval "exec vllm serve /mnt/models \ --served-model-name "facebook/opt-125m" "publishers/kserve-ci-e2e-test/models/facebook/opt-125m" \ --port 8000 \ ${ACCESS_LOG_ARGS} \ ${SHUTDOWN_TIMEOUT_ARGS} \ --enable-ssl-refresh \ --ssl-certfile /var/run/kserve/tls/tls.crt \ --ssl-keyfile /var/run/kserve/tls/tls.key \ ${VLLM_ADDITIONAL_ARGS} \ $@" - -- env: - name: HOME value: /home - name: VLLM_LOGGING_LEVEL value: DEBUG - name: VLLM_CPU_KVCACHE_SPACE value: "1" - name: VLLM_ENABLE_V1_MULTIPROCESSING value: "0" - name: USER value: nonroot - name: TORCHINDUCTOR_CACHE_DIR value: /tmp/torchinductor-cache - name: HF_HUB_CACHE value: /models image: public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:v0.19.0 imagePullPolicy: IfNotPresent lifecycle: preStop: exec: command: - /bin/sleep - "15" livenessProbe: failureThreshold: 10 httpGet: path: /health port: 8000 scheme: HTTPS periodSeconds: 10 successThreshold: 1 timeoutSeconds: 1 name: main ports: - containerPort: 8000 protocol: TCP readinessProbe: failureThreshold: 2 httpGet: path: /health port: 8000 scheme: HTTPS periodSeconds: 1 successThreshold: 1 timeoutSeconds: 1 resources: limits: cpu: "2" memory: 7Gi requests: cpu: 200m memory: 2Gi securityContext: allowPrivilegeEscalation: false capabilities: drop: - ALL readOnlyRootFilesystem: true 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 - mountPath: /mnt name: kserve-provision-location readOnly: true dnsPolicy: ClusterFirst initContainers: - args: - hf://facebook/opt-125m - /mnt/models - hf://edbeeching/opt-125m-lora - /mnt/lora/lora-adapter-1 env: - name: AWS_ACCESS_KEY_ID valueFrom: secretKeyRef: key: AWS_ACCESS_KEY_ID name: seaweedfs-s3-creds - name: AWS_SECRET_ACCESS_KEY valueFrom: secretKeyRef: key: AWS_SECRET_ACCESS_KEY name: seaweedfs-s3-creds - name: S3_USE_HTTPS value: "0" - name: S3_ENDPOINT value: s3-service.kserve:8333 - name: AWS_ENDPOINT_URL value: http://s3-service.kserve:8333 - name: S3_VERIFY_SSL value: "0" - name: AWS_CA_BUNDLE value: /etc/ssl/custom-certs/cabundle.crt - name: AWS_CA_BUNDLE_CONFIGMAP value: odh-kserve-custom-ca-bundle - name: HF_HUB_ENABLE_HF_TRANSFER value: "1" - name: HF_XET_HIGH_PERFORMANCE value: "1" - name: HF_XET_NUM_CONCURRENT_RANGE_GETS value: "8" - name: CA_BUNDLE_CONFIGMAP_NAME value: odh-kserve-custom-ca-bundle - name: CA_BUNDLE_VOLUME_MOUNT_POINT value: /etc/ssl/custom-certs image: quay.io/opendatahub/kserve-storage-initializer@sha256:33f82c067d520fe749f73749cf76b5935fe69bd9eb141504c6660698c7947743 imagePullPolicy: IfNotPresent name: storage-initializer resources: limits: cpu: "1" memory: 24Gi requests: cpu: 100m memory: 100Mi terminationMessagePath: /dev/termination-log terminationMessagePolicy: FallbackToLogsOnError volumeMounts: - mountPath: /mnt name: kserve-provision-location - mountPath: /etc/ssl/custom-certs name: cabundle-cert readOnly: true 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: llmisv3e414c2ba058a022dfd694dbcbac5b51-kserve-self-signed-certs - emptyDir: {} name: kserve-provision-location - configMap: defaultMode: 420 name: odh-kserve-custom-ca-bundle name: cabundle-cert status: availableReplicas: 1 conditions: - lastTransitionTime: "2026-07-02T21:42:41Z" lastUpdateTime: "2026-07-02T21:42:41Z" message: Deployment has minimum availability. reason: MinimumReplicasAvailable status: "True" type: Available - lastTransitionTime: "2026-07-02T21:41:10Z" lastUpdateTime: "2026-07-02T21:42:41Z" message: ReplicaSet "llmisvc-model-fb-opt-125m-with-7ca60146-kserve-cf76987fb" has successfully progressed. reason: NewReplicaSetAvailable status: "True" type: Progressing observedGeneration: 1 readyReplicas: 1 replicas: 1 updatedReplicas: 1 - apiVersion: apps/v1 kind: Deployment metadata: annotations: deployment.kubernetes.io/revision: "1" creationTimestamp: "2026-07-02T21:41:10Z" generation: 1 labels: app.kubernetes.io/component: llminferenceservice-router-scheduler app.kubernetes.io/name: llmisvc-model-fb-opt-125m-with-7ca60146 app.kubernetes.io/part-of: llminferenceservice 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:ownerReferences: .: {} k:{"uid":"9e661501-7bd8-4942-8472-a92b8c944b91"}: {} f:spec: f:progressDeadlineSeconds: {} f:replicas: {} f:revisionHistoryLimit: {} f:selector: {} f:strategy: f:type: {} f:template: f:metadata: f:annotations: .: {} f:app.kubernetes.io/version: {} f:certificates.kserve.io/expiration-v2: {} f:labels: .: {} f:app.kubernetes.io/component: {} f:app.kubernetes.io/name: {} f:app.kubernetes.io/part-of: {} f:spec: f:containers: k:{"name":"main"}: .: {} f:args: {} 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: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":"/tmp/tokenizer"}: .: {} f:mountPath: {} f:name: {} k:{"mountPath":"/var/run/kserve/tls"}: .: {} f:mountPath: {} f:name: {} f:readOnly: {} f:dnsPolicy: {} f:initContainers: .: {} k:{"name":"storage-initializer"}: .: {} f:args: {} f:env: .: {} k:{"name":"AWS_ACCESS_KEY_ID"}: .: {} f:name: {} f:valueFrom: .: {} f:secretKeyRef: {} k:{"name":"AWS_CA_BUNDLE"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_CA_BUNDLE_CONFIGMAP"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_ENDPOINT_URL"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_SECRET_ACCESS_KEY"}: .: {} f:name: {} f:valueFrom: .: {} f:secretKeyRef: {} k:{"name":"HF_HUB_ENABLE_HF_TRANSFER"}: .: {} f:name: {} f:value: {} k:{"name":"HF_XET_HIGH_PERFORMANCE"}: .: {} f:name: {} f:value: {} k:{"name":"HF_XET_NUM_CONCURRENT_RANGE_GETS"}: .: {} f:name: {} f:value: {} k:{"name":"S3_ENDPOINT"}: .: {} f:name: {} f:value: {} k:{"name":"S3_USE_HTTPS"}: .: {} f:name: {} f:value: {} k:{"name":"S3_VERIFY_SSL"}: .: {} f:name: {} f:value: {} k:{"name":"STORAGE_ALLOW_PATTERNS"}: .: {} f:name: {} f:value: {} f:image: {} f:imagePullPolicy: {} f:name: {} f:resources: .: {} f:limits: .: {} f:cpu: {} f:memory: {} f:requests: .: {} f:cpu: {} f:memory: {} f:terminationMessagePath: {} f:terminationMessagePolicy: {} f:volumeMounts: .: {} k:{"mountPath":"/mnt/models"}: .: {} f:mountPath: {} f:name: {} f:restartPolicy: {} f:schedulerName: {} f:securityContext: {} f:serviceAccount: {} f:serviceAccountName: {} f:terminationGracePeriodSeconds: {} f:volumes: .: {} k:{"name":"kserve-provision-location"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"tls-certs"}: .: {} f:name: {} f:secret: .: {} f:defaultMode: {} f:secretName: {} k:{"name":"tokenizer-cache"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"tokenizer-tmp"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"tokenizer-uds"}: .: {} f:emptyDir: {} f:name: {} manager: manager operation: Update time: "2026-07-02T21:41:10Z" - 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:updatedReplicas: {} manager: kube-controller-manager operation: Update subresource: status time: "2026-07-02T21:41:44Z" name: llmisvc-model-fb-opt-125m-with-7ca60146-kserve-router-scheduler namespace: kserve-ci-e2e-test ownerReferences: - apiVersion: serving.kserve.io/v1alpha2 blockOwnerDeletion: true controller: true kind: LLMInferenceService name: llmisvc-model-fb-opt-125m-with-7ca60146 uid: 9e661501-7bd8-4942-8472-a92b8c944b91 resourceVersion: "30959" uid: 76a04837-6905-4072-8c79-a3572e4168c3 spec: progressDeadlineSeconds: 600 replicas: 1 revisionHistoryLimit: 10 selector: matchLabels: app.kubernetes.io/component: llminferenceservice-router-scheduler app.kubernetes.io/name: llmisvc-model-fb-opt-125m-with-7ca60146 app.kubernetes.io/part-of: llminferenceservice strategy: type: Recreate template: metadata: annotations: app.kubernetes.io/version: 0.9.0 certificates.kserve.io/expiration-v2: "true" labels: app.kubernetes.io/component: llminferenceservice-router-scheduler app.kubernetes.io/name: llmisvc-model-fb-opt-125m-with-7ca60146 app.kubernetes.io/part-of: llminferenceservice spec: containers: - args: - --config-text - | apiVersion: inference.networking.x-k8s.io/v1alpha1 kind: EndpointPickerConfig plugins: - type: single-profile-handler - type: queue-scorer - type: prefix-cache-scorer - type: max-score-picker - type: lora-affinity-scorer schedulingProfiles: - name: default plugins: - pluginRef: lora-affinity-scorer weight: 4 - pluginRef: queue-scorer weight: 2 - pluginRef: prefix-cache-scorer weight: 3 - pluginRef: max-score-picker command: - /app/epp - --pool-name - llmisvc-model-fb-opt-125m-with-7ca60146-inference-pool - --pool-namespace - kserve-ci-e2e-test - --zap-encoder - json - --grpc-port - "9002" - --grpc-health-port - "9003" - --enable-cert-reload=true - --secure-serving=true - --model-server-metrics-scheme=https - --cert-path=/var/run/kserve/tls env: - name: SSL_CERT_DIR value: /var/run/kserve/tls:/var/run/secrets/kubernetes.io/serviceaccount:/etc/pki/tls/certs image: ghcr.io/llm-d/llm-d-router-endpoint-picker:v0.9.0-rc.2 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: requests: cpu: 256m memory: 500Mi 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 - mountPath: /tmp/tokenizer name: tokenizer-uds dnsPolicy: ClusterFirst initContainers: - args: - hf://facebook/opt-125m - /mnt/models env: - name: AWS_ACCESS_KEY_ID valueFrom: secretKeyRef: key: AWS_ACCESS_KEY_ID name: seaweedfs-s3-creds - name: AWS_SECRET_ACCESS_KEY valueFrom: secretKeyRef: key: AWS_SECRET_ACCESS_KEY name: seaweedfs-s3-creds - name: S3_USE_HTTPS value: "0" - name: S3_ENDPOINT value: s3-service.kserve:8333 - name: AWS_ENDPOINT_URL value: http://s3-service.kserve:8333 - name: S3_VERIFY_SSL value: "0" - name: AWS_CA_BUNDLE value: /etc/ssl/custom-certs/cabundle.crt - name: AWS_CA_BUNDLE_CONFIGMAP value: odh-kserve-custom-ca-bundle - name: HF_HUB_ENABLE_HF_TRANSFER value: "1" - name: HF_XET_HIGH_PERFORMANCE value: "1" - name: HF_XET_NUM_CONCURRENT_RANGE_GETS value: "8" - name: STORAGE_ALLOW_PATTERNS value: '["tokenizer.json", "tokenizer_config.json", "special_tokens_map.json", "vocab.json", "merges.txt", "config.json", "generation_config.json"]' image: quay.io/opendatahub/kserve-storage-initializer@sha256:33f82c067d520fe749f73749cf76b5935fe69bd9eb141504c6660698c7947743 imagePullPolicy: IfNotPresent name: storage-initializer resources: limits: cpu: "1" memory: 24Gi requests: cpu: 100m memory: 100Mi terminationMessagePath: /dev/termination-log terminationMessagePolicy: FallbackToLogsOnError volumeMounts: - mountPath: /mnt/models name: kserve-provision-location restartPolicy: Always schedulerName: default-scheduler securityContext: {} serviceAccount: llmisvc-model-fb-opt-125m-with-7ca60146-epp-sa serviceAccountName: llmisvc-model-fb-opt-125m-with-7ca60146-epp-sa terminationGracePeriodSeconds: 60 volumes: - name: tls-certs secret: defaultMode: 420 secretName: llmisv3e414c2ba058a022dfd694dbcbac5b51-kserve-self-signed-certs - emptyDir: {} name: tokenizer-uds - emptyDir: {} name: tokenizer-tmp - emptyDir: {} name: tokenizer-cache - emptyDir: {} name: kserve-provision-location status: availableReplicas: 1 conditions: - lastTransitionTime: "2026-07-02T21:41:44Z" lastUpdateTime: "2026-07-02T21:41:44Z" message: Deployment has minimum availability. reason: MinimumReplicasAvailable status: "True" type: Available - lastTransitionTime: "2026-07-02T21:41:10Z" lastUpdateTime: "2026-07-02T21:41:44Z" message: ReplicaSet "llmisvc-model-fb-opt-125m-with-7ca60146-kserve-router-scheduler-766bf6cdf7" has successfully progressed. reason: NewReplicaSetAvailable status: "True" type: Progressing observedGeneration: 1 readyReplicas: 1 replicas: 1 updatedReplicas: 1 - apiVersion: apps/v1 kind: Deployment metadata: annotations: deployment.kubernetes.io/revision: "1" creationTimestamp: "2026-07-02T21:57:59Z" generation: 1 labels: app.kubernetes.io/component: llminferenceservice-workload app.kubernetes.io/name: llmisvc-model-fb-opt-125m-with-ba4d693a 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":"d427c7a4-20aa-4e71-811e-24adfd64d475"}: {} 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":"TORCHINDUCTOR_CACHE_DIR"}: .: {} f:name: {} f:value: {} k:{"name":"USER"}: .: {} f:name: {} f:value: {} k:{"name":"VLLM_CPU_KVCACHE_SPACE"}: .: {} f:name: {} f:value: {} k:{"name":"VLLM_ENABLE_V1_MULTIPROCESSING"}: .: {} f:name: {} f:value: {} 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: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":"/mnt"}: .: {} f:mountPath: {} f:name: {} f:readOnly: {} 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:initContainers: .: {} k:{"name":"storage-initializer"}: .: {} f:args: {} f:env: .: {} k:{"name":"AWS_ACCESS_KEY_ID"}: .: {} f:name: {} f:valueFrom: .: {} f:secretKeyRef: {} k:{"name":"AWS_CA_BUNDLE"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_CA_BUNDLE_CONFIGMAP"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_ENDPOINT_URL"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_SECRET_ACCESS_KEY"}: .: {} f:name: {} f:valueFrom: .: {} f:secretKeyRef: {} k:{"name":"CA_BUNDLE_CONFIGMAP_NAME"}: .: {} f:name: {} f:value: {} k:{"name":"CA_BUNDLE_VOLUME_MOUNT_POINT"}: .: {} f:name: {} f:value: {} k:{"name":"HF_HUB_ENABLE_HF_TRANSFER"}: .: {} f:name: {} f:value: {} k:{"name":"HF_XET_HIGH_PERFORMANCE"}: .: {} f:name: {} f:value: {} k:{"name":"HF_XET_NUM_CONCURRENT_RANGE_GETS"}: .: {} f:name: {} f:value: {} k:{"name":"S3_ENDPOINT"}: .: {} f:name: {} f:value: {} k:{"name":"S3_USE_HTTPS"}: .: {} f:name: {} f:value: {} k:{"name":"S3_VERIFY_SSL"}: .: {} f:name: {} f:value: {} f:image: {} f:imagePullPolicy: {} f:name: {} f:resources: .: {} f:limits: .: {} f:cpu: {} f:memory: {} f:requests: .: {} f:cpu: {} f:memory: {} f:terminationMessagePath: {} f:terminationMessagePolicy: {} f:volumeMounts: .: {} k:{"mountPath":"/etc/ssl/custom-certs"}: .: {} f:mountPath: {} f:name: {} f:readOnly: {} k:{"mountPath":"/mnt"}: .: {} f:mountPath: {} f:name: {} f:restartPolicy: {} f:schedulerName: {} f:securityContext: {} f:terminationGracePeriodSeconds: {} f:volumes: .: {} k:{"name":"cabundle-cert"}: .: {} f:configMap: .: {} f:defaultMode: {} f:name: {} f:name: {} k:{"name":"dshm"}: .: {} f:emptyDir: .: {} f:medium: {} f:sizeLimit: {} f:name: {} k:{"name":"home"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"kserve-provision-location"}: .: {} 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: "2026-07-02T21:57:59Z" - 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:updatedReplicas: {} manager: kube-controller-manager operation: Update subresource: status time: "2026-07-02T21:59:20Z" name: llmisvc-model-fb-opt-125m-with-ba4d693a-kserve namespace: kserve-ci-e2e-test ownerReferences: - apiVersion: serving.kserve.io/v1alpha2 blockOwnerDeletion: true controller: true kind: LLMInferenceService name: llmisvc-model-fb-opt-125m-with-ba4d693a uid: d427c7a4-20aa-4e71-811e-24adfd64d475 resourceVersion: "49328" uid: 5c2fa809-7b07-4e3f-af9c-fd2cf1189d03 spec: progressDeadlineSeconds: 600 replicas: 1 revisionHistoryLimit: 10 selector: matchLabels: app.kubernetes.io/component: llminferenceservice-workload app.kubernetes.io/name: llmisvc-model-fb-opt-125m-with-ba4d693a 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: llmisvc-model-fb-opt-125m-with-ba4d693a app.kubernetes.io/part-of: llminferenceservice kserve.io/component: workload llm-d.ai/role: both spec: containers: - args: - --enable-lora - --lora-modules - '''{"name":"lora-adapter-1","path":"/mnt/lora/lora-adapter-1"}''' - '''{"name":"publishers/kserve-ci-e2e-test/models/lora-adapter-1","path":"/mnt/lora/lora-adapter-1"}''' command: - /bin/bash - -c - |- if [ -f /etc/profile.d/ibm-aiu-setup.sh ]; then source /etc/profile.d/ibm-aiu-setup.sh fi if [ "$KSERVE_INFER_ROCE" = "true" ]; then echo "Trying to infer RoCE configs ... " grep -H . /sys/class/infiniband/*/ports/*/gids/* 2>/dev/null grep -H . /sys/class/infiniband/*/ports/*/gid_attrs/types/* 2>/dev/null cat /proc/driver/nvidia/params KSERVE_INFER_IB_GID_INDEX_GREP=${KSERVE_INFER_IB_GID_INDEX_GREP:-"RoCE v2"} echo "[Infer RoCE] Discovering active HCAs ..." active_hcas=() # Loop through all mlx5 devices found in sysfs for hca_dir in /sys/class/infiniband/mlx5_*; do # Ensure it's a directory before proceeding if [ -d "$hca_dir" ]; then hca_name=$(basename "$hca_dir") port_state_file="$hca_dir/ports/1/state" # Assume port 1 type_file="$hca_dir/ports/1/gid_attrs/types/*" echo "[Infer RoCE] Check if the port state file ${port_state_file} exists and contains 'ACTIVE'" if [ -f "$port_state_file" ] && grep -q "ACTIVE" "$port_state_file" && grep -q "${KSERVE_INFER_IB_GID_INDEX_GREP}" ${type_file} 2>/dev/null; then echo "[Infer RoCE] Found active HCA: $hca_name" active_hcas+=("$hca_name") else echo "[Infer RoCE] Skipping inactive or down HCA: $hca_name" fi fi done # Check if we found any active HCAs if [ ${#active_hcas[@]} -gt 0 ]; then # Join the array elements with a comma hca_port_pairs=() for hca in "${active_hcas[@]}"; do hca_port_pairs+=("${hca}:1") done active_hca_list=$(IFS=,; echo "${active_hcas[*]}") hca_port_pairs_list=$(IFS=,; echo "${hca_port_pairs[*]}") echo "[Infer RoCE] Setting active HCAs: ${active_hca_list}" export NCCL_IB_HCA=${NCCL_IB_HCA:-${active_hca_list}} export NVSHMEM_HCA_LIST=${NVSHMEM_HCA_LIST:-${hca_port_pairs_list}} export UCX_NET_DEVICES=${UCX_NET_DEVICES:-${hca_port_pairs_list}} echo "[Infer RoCE] NCCL_IB_HCA=${NCCL_IB_HCA}" echo "[Infer RoCE] NVSHMEM_HCA_LIST=${NVSHMEM_HCA_LIST}" echo "[Infer RoCE] UCX_NET_DEVICES=${UCX_NET_DEVICES}" else echo "[Infer RoCE] WARNING: No active RoCE HCAs found. NCCL_IB_HCA will not be set." fi if [ ${#active_hcas[@]} -gt 0 ]; then echo "[Infer RoCE] Finding GID_INDEX for each active HCA (SR-IOV compatible)..." # For SR-IOV environments, find the most common IPv4 RoCE v2 GID index across all HCAs declare -A gid_index_count declare -A hca_gid_index for hca_name in "${active_hcas[@]}"; do echo "[Infer RoCE] Processing HCA: ${hca_name}" # Find all RoCE v2 IPv4 GIDs for this HCA and count by index for tpath in /sys/class/infiniband/${hca_name}/ports/1/gid_attrs/types/*; do if grep -q "${KSERVE_INFER_IB_GID_INDEX_GREP}" "$tpath" 2>/dev/null; then idx=$(basename "$tpath") gid_file="/sys/class/infiniband/${hca_name}/ports/1/gids/${idx}" # Check for IPv4 GID (contains ffff:) if [ -f "$gid_file" ] && grep -q "ffff:" "$gid_file"; then gid_value=$(cat "$gid_file" 2>/dev/null || echo "") echo "[Infer RoCE] Found IPv4 RoCE v2 GID for ${hca_name}: index=${idx}, gid=${gid_value}" hca_gid_index["${hca_name}"]="${idx}" gid_index_count["${idx}"]=$((${gid_index_count["${idx}"]} + 1)) break # Use first found IPv4 GID per HCA fi fi done done # Find the most common GID index (most likely to be consistent across nodes) best_gid_index="" max_count=0 for idx in "${!gid_index_count[@]}"; do count=${gid_index_count["${idx}"]} echo "[Infer RoCE] GID_INDEX ${idx} found on ${count} HCAs" if [ $count -gt $max_count ]; then max_count=$count best_gid_index="$idx" fi done # Use deterministic fallback if tied - prefer index 3 (SR-IOV standard) if [ ${#gid_index_count[@]} -gt 1 ]; then echo "[Infer RoCE] Multiple GID indices found, selecting most common: ${best_gid_index}" # If there's a tie, prefer index 3 as it's most common in SR-IOV setups if [ -n "${gid_index_count['3']}" ] && [ "${gid_index_count['3']}" -eq "$max_count" ]; then best_gid_index="3" echo "[Infer RoCE] Using deterministic fallback: GID_INDEX=3 (SR-IOV standard)" fi fi # Check if GID_INDEX is already set via environment variables if [ -n "${NCCL_IB_GID_INDEX}" ]; then echo "[Infer RoCE] Using pre-configured NCCL_IB_GID_INDEX=${NCCL_IB_GID_INDEX} from environment" export NVSHMEM_IB_GID_INDEX=${NVSHMEM_IB_GID_INDEX:-$NCCL_IB_GID_INDEX} export UCX_IB_GID_INDEX=${UCX_IB_GID_INDEX:-$NCCL_IB_GID_INDEX} echo "[Infer RoCE] Using pre-configured GID_INDEX=${NCCL_IB_GID_INDEX} for NCCL, NVSHMEM, and UCX" elif [ -n "$best_gid_index" ]; then echo "[Infer RoCE] Selected GID_INDEX: ${best_gid_index} (found on ${max_count} HCAs)" export NCCL_IB_GID_INDEX=${NCCL_IB_GID_INDEX:-$best_gid_index} export NVSHMEM_IB_GID_INDEX=${NVSHMEM_IB_GID_INDEX:-$best_gid_index} export UCX_IB_GID_INDEX=${UCX_IB_GID_INDEX:-$best_gid_index} echo "[Infer RoCE] Exported GID_INDEX=${best_gid_index} for NCCL, NVSHMEM, and UCX" else echo "[Infer RoCE] ERROR: No valid IPv4 ${KSERVE_INFER_IB_GID_INDEX_GREP} GID_INDEX found on any HCA." fi else echo "[Infer RoCE] No active HCAs found, skipping GID_INDEX inference." fi fi # --disable-access-log-for-endpoints landed in vLLM 0.16.0 (vllm-project/vllm#30011). # Older versions still need the blanket --disable-uvicorn-access-log. ACCESS_LOG_ARGS="--disable-uvicorn-access-log" VLLM_VERSION=$(vllm --version 2>/dev/null | tail -1 | awk '{print $NF}') echo "[access-log-detect] vllm version='${VLLM_VERSION}'" if [[ "$VLLM_VERSION" =~ ^[0-9]+\.[0-9]+ ]] && [ "$(printf '%s\n%s\n' "0.16.0" "${VLLM_VERSION}" | sort -V | head -1)" = "0.16.0" ]; then ACCESS_LOG_ARGS="--disable-access-log-for-endpoints /health,/metrics,/ping" fi echo "[access-log-detect] selected ACCESS_LOG_ARGS='${ACCESS_LOG_ARGS}'" # --shutdown-timeout landed in vLLM 0.18.0 (vllm-project/vllm#36666). SHUTDOWN_TIMEOUT_ARGS="" if [[ "$VLLM_VERSION" =~ ^[0-9]+\.[0-9]+ ]] && [ "$(printf '%s\n%s\n' "0.18.0" "${VLLM_VERSION}" | sort -V | head -1)" = "0.18.0" ]; then SHUTDOWN_TIMEOUT_ARGS="--shutdown-timeout 40" fi eval "exec vllm serve /mnt/models \ --served-model-name "facebook/opt-125m" "publishers/kserve-ci-e2e-test/models/facebook/opt-125m" \ --port 8000 \ ${ACCESS_LOG_ARGS} \ ${SHUTDOWN_TIMEOUT_ARGS} \ --enable-ssl-refresh \ --ssl-certfile /var/run/kserve/tls/tls.crt \ --ssl-keyfile /var/run/kserve/tls/tls.key \ ${VLLM_ADDITIONAL_ARGS} \ $@" - -- env: - name: HOME value: /home - name: VLLM_LOGGING_LEVEL value: DEBUG - name: VLLM_CPU_KVCACHE_SPACE value: "1" - name: VLLM_ENABLE_V1_MULTIPROCESSING value: "0" - name: USER value: nonroot - name: TORCHINDUCTOR_CACHE_DIR value: /tmp/torchinductor-cache - name: HF_HUB_CACHE value: /models image: public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:v0.19.0 imagePullPolicy: IfNotPresent lifecycle: preStop: exec: command: - /bin/sleep - "15" livenessProbe: failureThreshold: 10 httpGet: path: /health port: 8000 scheme: HTTPS periodSeconds: 10 successThreshold: 1 timeoutSeconds: 1 name: main ports: - containerPort: 8000 protocol: TCP readinessProbe: failureThreshold: 2 httpGet: path: /health port: 8000 scheme: HTTPS periodSeconds: 1 successThreshold: 1 timeoutSeconds: 1 resources: limits: cpu: "2" memory: 7Gi requests: cpu: 200m memory: 2Gi securityContext: allowPrivilegeEscalation: false capabilities: drop: - ALL readOnlyRootFilesystem: true 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 - mountPath: /mnt name: kserve-provision-location readOnly: true dnsPolicy: ClusterFirst initContainers: - args: - hf://facebook/opt-125m - /mnt/models - hf://edbeeching/opt-125m-lora - /mnt/lora/lora-adapter-1 env: - name: AWS_ACCESS_KEY_ID valueFrom: secretKeyRef: key: AWS_ACCESS_KEY_ID name: seaweedfs-s3-creds - name: AWS_SECRET_ACCESS_KEY valueFrom: secretKeyRef: key: AWS_SECRET_ACCESS_KEY name: seaweedfs-s3-creds - name: S3_USE_HTTPS value: "0" - name: S3_ENDPOINT value: s3-service.kserve:8333 - name: AWS_ENDPOINT_URL value: http://s3-service.kserve:8333 - name: S3_VERIFY_SSL value: "0" - name: AWS_CA_BUNDLE value: /etc/ssl/custom-certs/cabundle.crt - name: AWS_CA_BUNDLE_CONFIGMAP value: odh-kserve-custom-ca-bundle - name: HF_HUB_ENABLE_HF_TRANSFER value: "1" - name: HF_XET_HIGH_PERFORMANCE value: "1" - name: HF_XET_NUM_CONCURRENT_RANGE_GETS value: "8" - name: CA_BUNDLE_CONFIGMAP_NAME value: odh-kserve-custom-ca-bundle - name: CA_BUNDLE_VOLUME_MOUNT_POINT value: /etc/ssl/custom-certs image: quay.io/opendatahub/kserve-storage-initializer@sha256:33f82c067d520fe749f73749cf76b5935fe69bd9eb141504c6660698c7947743 imagePullPolicy: IfNotPresent name: storage-initializer resources: limits: cpu: "1" memory: 24Gi requests: cpu: 100m memory: 100Mi terminationMessagePath: /dev/termination-log terminationMessagePolicy: FallbackToLogsOnError volumeMounts: - mountPath: /mnt name: kserve-provision-location - mountPath: /etc/ssl/custom-certs name: cabundle-cert readOnly: true 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: llmisv77ff2528d3e9b4972cd9335229fce9f0-kserve-self-signed-certs - emptyDir: {} name: kserve-provision-location - configMap: defaultMode: 420 name: odh-kserve-custom-ca-bundle name: cabundle-cert status: availableReplicas: 1 conditions: - lastTransitionTime: "2026-07-02T21:59:20Z" lastUpdateTime: "2026-07-02T21:59:20Z" message: Deployment has minimum availability. reason: MinimumReplicasAvailable status: "True" type: Available - lastTransitionTime: "2026-07-02T21:57:59Z" lastUpdateTime: "2026-07-02T21:59:20Z" message: ReplicaSet "llmisvc-model-fb-opt-125m-with-ba4d693a-kserve-598d446855" has successfully progressed. reason: NewReplicaSetAvailable status: "True" type: Progressing observedGeneration: 1 readyReplicas: 1 replicas: 1 updatedReplicas: 1 - apiVersion: apps/v1 kind: Deployment metadata: annotations: deployment.kubernetes.io/revision: "1" creationTimestamp: "2026-07-02T21:57:59Z" generation: 1 labels: app.kubernetes.io/component: llminferenceservice-router-scheduler app.kubernetes.io/name: llmisvc-model-fb-opt-125m-with-ba4d693a app.kubernetes.io/part-of: llminferenceservice 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:ownerReferences: .: {} k:{"uid":"d427c7a4-20aa-4e71-811e-24adfd64d475"}: {} f:spec: f:progressDeadlineSeconds: {} f:replicas: {} f:revisionHistoryLimit: {} f:selector: {} f:strategy: f:type: {} f:template: f:metadata: f:annotations: .: {} f:app.kubernetes.io/version: {} f:certificates.kserve.io/expiration-v2: {} f:labels: .: {} f:app.kubernetes.io/component: {} f:app.kubernetes.io/name: {} f:app.kubernetes.io/part-of: {} f:spec: f:containers: k:{"name":"main"}: .: {} f:args: {} 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: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":"/tmp/tokenizer"}: .: {} f:mountPath: {} f:name: {} k:{"mountPath":"/var/run/kserve/tls"}: .: {} f:mountPath: {} f:name: {} f:readOnly: {} f:dnsPolicy: {} f:initContainers: .: {} k:{"name":"storage-initializer"}: .: {} f:args: {} f:env: .: {} k:{"name":"AWS_ACCESS_KEY_ID"}: .: {} f:name: {} f:valueFrom: .: {} f:secretKeyRef: {} k:{"name":"AWS_CA_BUNDLE"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_CA_BUNDLE_CONFIGMAP"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_ENDPOINT_URL"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_SECRET_ACCESS_KEY"}: .: {} f:name: {} f:valueFrom: .: {} f:secretKeyRef: {} k:{"name":"HF_HUB_ENABLE_HF_TRANSFER"}: .: {} f:name: {} f:value: {} k:{"name":"HF_XET_HIGH_PERFORMANCE"}: .: {} f:name: {} f:value: {} k:{"name":"HF_XET_NUM_CONCURRENT_RANGE_GETS"}: .: {} f:name: {} f:value: {} k:{"name":"S3_ENDPOINT"}: .: {} f:name: {} f:value: {} k:{"name":"S3_USE_HTTPS"}: .: {} f:name: {} f:value: {} k:{"name":"S3_VERIFY_SSL"}: .: {} f:name: {} f:value: {} k:{"name":"STORAGE_ALLOW_PATTERNS"}: .: {} f:name: {} f:value: {} f:image: {} f:imagePullPolicy: {} f:name: {} f:resources: .: {} f:limits: .: {} f:cpu: {} f:memory: {} f:requests: .: {} f:cpu: {} f:memory: {} f:terminationMessagePath: {} f:terminationMessagePolicy: {} f:volumeMounts: .: {} k:{"mountPath":"/mnt/models"}: .: {} f:mountPath: {} f:name: {} f:restartPolicy: {} f:schedulerName: {} f:securityContext: {} f:serviceAccount: {} f:serviceAccountName: {} f:terminationGracePeriodSeconds: {} f:volumes: .: {} k:{"name":"kserve-provision-location"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"tls-certs"}: .: {} f:name: {} f:secret: .: {} f:defaultMode: {} f:secretName: {} k:{"name":"tokenizer-cache"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"tokenizer-tmp"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"tokenizer-uds"}: .: {} f:emptyDir: {} f:name: {} manager: manager operation: Update time: "2026-07-02T21:57:59Z" - 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:updatedReplicas: {} manager: kube-controller-manager operation: Update subresource: status time: "2026-07-02T21:58:34Z" name: llmisvc-model-fb-opt-125m-with-ba4d693a-kserve-router-scheduler namespace: kserve-ci-e2e-test ownerReferences: - apiVersion: serving.kserve.io/v1alpha2 blockOwnerDeletion: true controller: true kind: LLMInferenceService name: llmisvc-model-fb-opt-125m-with-ba4d693a uid: d427c7a4-20aa-4e71-811e-24adfd64d475 resourceVersion: "48627" uid: 480afe83-6e53-40c0-81e0-a6f197f683a0 spec: progressDeadlineSeconds: 600 replicas: 1 revisionHistoryLimit: 10 selector: matchLabels: app.kubernetes.io/component: llminferenceservice-router-scheduler app.kubernetes.io/name: llmisvc-model-fb-opt-125m-with-ba4d693a app.kubernetes.io/part-of: llminferenceservice strategy: type: Recreate template: metadata: annotations: app.kubernetes.io/version: 0.9.0 certificates.kserve.io/expiration-v2: "true" labels: app.kubernetes.io/component: llminferenceservice-router-scheduler app.kubernetes.io/name: llmisvc-model-fb-opt-125m-with-ba4d693a app.kubernetes.io/part-of: llminferenceservice spec: containers: - args: - --config-text - | apiVersion: inference.networking.x-k8s.io/v1alpha1 kind: EndpointPickerConfig plugins: - type: single-profile-handler - type: queue-scorer - type: prefix-cache-scorer - type: max-score-picker - type: lora-affinity-scorer schedulingProfiles: - name: default plugins: - pluginRef: lora-affinity-scorer weight: 4 - pluginRef: queue-scorer weight: 2 - pluginRef: prefix-cache-scorer weight: 3 - pluginRef: max-score-picker command: - /app/epp - --pool-name - llmisvc-model-fb-opt-125m-with-ba4d693a-inference-pool - --pool-namespace - kserve-ci-e2e-test - --zap-encoder - json - --grpc-port - "9002" - --grpc-health-port - "9003" - --enable-cert-reload=true - --secure-serving=true - --model-server-metrics-scheme=https - --cert-path=/var/run/kserve/tls env: - name: SSL_CERT_DIR value: /var/run/kserve/tls:/var/run/secrets/kubernetes.io/serviceaccount:/etc/pki/tls/certs image: ghcr.io/llm-d/llm-d-router-endpoint-picker:v0.9.0-rc.2 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: requests: cpu: 256m memory: 500Mi 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 - mountPath: /tmp/tokenizer name: tokenizer-uds dnsPolicy: ClusterFirst initContainers: - args: - hf://facebook/opt-125m - /mnt/models env: - name: AWS_ACCESS_KEY_ID valueFrom: secretKeyRef: key: AWS_ACCESS_KEY_ID name: seaweedfs-s3-creds - name: AWS_SECRET_ACCESS_KEY valueFrom: secretKeyRef: key: AWS_SECRET_ACCESS_KEY name: seaweedfs-s3-creds - name: S3_USE_HTTPS value: "0" - name: S3_ENDPOINT value: s3-service.kserve:8333 - name: AWS_ENDPOINT_URL value: http://s3-service.kserve:8333 - name: S3_VERIFY_SSL value: "0" - name: AWS_CA_BUNDLE value: /etc/ssl/custom-certs/cabundle.crt - name: AWS_CA_BUNDLE_CONFIGMAP value: odh-kserve-custom-ca-bundle - name: HF_HUB_ENABLE_HF_TRANSFER value: "1" - name: HF_XET_HIGH_PERFORMANCE value: "1" - name: HF_XET_NUM_CONCURRENT_RANGE_GETS value: "8" - name: STORAGE_ALLOW_PATTERNS value: '["tokenizer.json", "tokenizer_config.json", "special_tokens_map.json", "vocab.json", "merges.txt", "config.json", "generation_config.json"]' image: quay.io/opendatahub/kserve-storage-initializer@sha256:33f82c067d520fe749f73749cf76b5935fe69bd9eb141504c6660698c7947743 imagePullPolicy: IfNotPresent name: storage-initializer resources: limits: cpu: "1" memory: 24Gi requests: cpu: 100m memory: 100Mi terminationMessagePath: /dev/termination-log terminationMessagePolicy: FallbackToLogsOnError volumeMounts: - mountPath: /mnt/models name: kserve-provision-location restartPolicy: Always schedulerName: default-scheduler securityContext: {} serviceAccount: llmisvc-model-fb-opt-125m-with-ba4d693a-epp-sa serviceAccountName: llmisvc-model-fb-opt-125m-with-ba4d693a-epp-sa terminationGracePeriodSeconds: 60 volumes: - name: tls-certs secret: defaultMode: 420 secretName: llmisv77ff2528d3e9b4972cd9335229fce9f0-kserve-self-signed-certs - emptyDir: {} name: tokenizer-uds - emptyDir: {} name: tokenizer-tmp - emptyDir: {} name: tokenizer-cache - emptyDir: {} name: kserve-provision-location status: availableReplicas: 1 conditions: - lastTransitionTime: "2026-07-02T21:58:34Z" lastUpdateTime: "2026-07-02T21:58:34Z" message: Deployment has minimum availability. reason: MinimumReplicasAvailable status: "True" type: Available - lastTransitionTime: "2026-07-02T21:57:59Z" lastUpdateTime: "2026-07-02T21:58:34Z" message: ReplicaSet "llmisvc-model-fb-opt-125m-with-ba4d693a-kserve-router-scheduler-67cc76c46d" has successfully progressed. reason: NewReplicaSetAvailable status: "True" type: Progressing observedGeneration: 1 readyReplicas: 1 replicas: 1 updatedReplicas: 1 - apiVersion: apps/v1 kind: Deployment metadata: annotations: deployment.kubernetes.io/revision: "1" creationTimestamp: "2026-07-02T22:08:28Z" generation: 1 labels: app.kubernetes.io/component: llminferenceservice-workload app.kubernetes.io/name: llmisvc-model-pvc-router-manage-e8706282 app.kubernetes.io/part-of: llminferenceservice kserve.io/component: workload llm-d.ai/role: decode 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":"9f190595-570c-42e2-b950-44b09c28e1f5"}: {} 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:command: {} f:env: .: {} k:{"name":"HF_HUB_CACHE"}: .: {} f:name: {} f:value: {} k:{"name":"HOME"}: .: {} f:name: {} f:value: {} k:{"name":"TORCHINDUCTOR_CACHE_DIR"}: .: {} f:name: {} f:value: {} k:{"name":"USER"}: .: {} f:name: {} f:value: {} k:{"name":"VLLM_CPU_KVCACHE_SPACE"}: .: {} f:name: {} f:value: {} k:{"name":"VLLM_ENABLE_V1_MULTIPROCESSING"}: .: {} f:name: {} f:value: {} 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:initialDelaySeconds: {} f:periodSeconds: {} f:successThreshold: {} f:timeoutSeconds: {} f:name: {} f:ports: .: {} k:{"containerPort":8001,"protocol":"TCP"}: .: {} f:containerPort: {} f:protocol: {} f:readinessProbe: .: {} f:failureThreshold: {} f:httpGet: .: {} f:path: {} f:port: {} f:scheme: {} 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: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":"/mnt/models"}: .: {} f:mountPath: {} f:name: {} f:readOnly: {} 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:initContainers: .: {} k:{"name":"llm-d-routing-sidecar"}: .: {} f:command: {} f:env: .: {} k:{"name":"INFERENCE_POOL_NAME"}: .: {} f:name: {} f:value: {} k:{"name":"INFERENCE_POOL_NAMESPACE"}: .: {} f:name: {} f:valueFrom: .: {} f:fieldRef: {} k:{"name":"SSL_CERT_DIR"}: .: {} f:name: {} f:value: {} f:image: {} f:imagePullPolicy: {} f:livenessProbe: .: {} f:failureThreshold: {} f:httpGet: .: {} f:path: {} f:port: {} f:scheme: {} f:initialDelaySeconds: {} f:periodSeconds: {} f:successThreshold: {} f:timeoutSeconds: {} f:name: {} f:ports: .: {} k:{"containerPort":8000,"protocol":"TCP"}: .: {} f:containerPort: {} f:protocol: {} f:readinessProbe: .: {} f:failureThreshold: {} f:httpGet: .: {} f:path: {} f:port: {} f:scheme: {} f:initialDelaySeconds: {} f:periodSeconds: {} f:successThreshold: {} f:timeoutSeconds: {} f:resources: {} f:restartPolicy: {} f:securityContext: .: {} f:allowPrivilegeEscalation: {} f:capabilities: .: {} f:drop: {} f:readOnlyRootFilesystem: {} f:runAsNonRoot: {} f:terminationMessagePath: {} f:terminationMessagePolicy: {} f:volumeMounts: .: {} k:{"mountPath":"/var/run/kserve/tls"}: .: {} f:mountPath: {} f:name: {} f:readOnly: {} f:restartPolicy: {} f:schedulerName: {} f:securityContext: {} f:serviceAccount: {} f:serviceAccountName: {} f:terminationGracePeriodSeconds: {} f:volumes: .: {} k:{"name":"dshm"}: .: {} f:emptyDir: .: {} f:medium: {} f:sizeLimit: {} f:name: {} k:{"name":"home"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"kserve-pvc-source"}: .: {} f:name: {} f:persistentVolumeClaim: .: {} f:claimName: {} 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: "2026-07-02T22:08:28Z" - 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:updatedReplicas: {} manager: kube-controller-manager operation: Update subresource: status time: "2026-07-02T22:11:31Z" name: llmisvc-model-pvc-router-manage-e8706282-kserve namespace: kserve-ci-e2e-test ownerReferences: - apiVersion: serving.kserve.io/v1alpha2 blockOwnerDeletion: true controller: true kind: LLMInferenceService name: llmisvc-model-pvc-router-manage-e8706282 uid: 9f190595-570c-42e2-b950-44b09c28e1f5 resourceVersion: "60813" uid: 588e264d-b36c-45c5-8e8d-02a6a771a190 spec: progressDeadlineSeconds: 600 replicas: 1 revisionHistoryLimit: 10 selector: matchLabels: app.kubernetes.io/component: llminferenceservice-workload app.kubernetes.io/name: llmisvc-model-pvc-router-manage-e8706282 app.kubernetes.io/part-of: llminferenceservice kserve.io/component: workload llm-d.ai/role: decode strategy: rollingUpdate: maxSurge: 25% maxUnavailable: 25% type: RollingUpdate template: metadata: labels: app.kubernetes.io/component: llminferenceservice-workload app.kubernetes.io/name: llmisvc-model-pvc-router-manage-e8706282 app.kubernetes.io/part-of: llminferenceservice kserve.io/component: workload llm-d.ai/role: decode spec: containers: - command: - /bin/bash - -c - |- if [ -f /etc/profile.d/ibm-aiu-setup.sh ]; then source /etc/profile.d/ibm-aiu-setup.sh fi if [ "$KSERVE_INFER_ROCE" = "true" ]; then echo "Trying to infer RoCE configs ... " grep -H . /sys/class/infiniband/*/ports/*/gids/* 2>/dev/null grep -H . /sys/class/infiniband/*/ports/*/gid_attrs/types/* 2>/dev/null cat /proc/driver/nvidia/params KSERVE_INFER_IB_GID_INDEX_GREP=${KSERVE_INFER_IB_GID_INDEX_GREP:-"RoCE v2"} echo "[Infer RoCE] Discovering active HCAs ..." active_hcas=() # Loop through all mlx5 devices found in sysfs for hca_dir in /sys/class/infiniband/mlx5_*; do # Ensure it's a directory before proceeding if [ -d "$hca_dir" ]; then hca_name=$(basename "$hca_dir") port_state_file="$hca_dir/ports/1/state" # Assume port 1 type_file="$hca_dir/ports/1/gid_attrs/types/*" echo "[Infer RoCE] Check if the port state file ${port_state_file} exists and contains 'ACTIVE'" if [ -f "$port_state_file" ] && grep -q "ACTIVE" "$port_state_file" && grep -q "${KSERVE_INFER_IB_GID_INDEX_GREP}" ${type_file} 2>/dev/null; then echo "[Infer RoCE] Found active HCA: $hca_name" active_hcas+=("$hca_name") else echo "[Infer RoCE] Skipping inactive or down HCA: $hca_name" fi fi done # Check if we found any active HCAs if [ ${#active_hcas[@]} -gt 0 ]; then # Join the array elements with a comma hca_port_pairs=() for hca in "${active_hcas[@]}"; do hca_port_pairs+=("${hca}:1") done active_hca_list=$(IFS=,; echo "${active_hcas[*]}") hca_port_pairs_list=$(IFS=,; echo "${hca_port_pairs[*]}") echo "[Infer RoCE] Setting active HCAs: ${active_hca_list}" export NCCL_IB_HCA=${NCCL_IB_HCA:-${active_hca_list}} export NVSHMEM_HCA_LIST=${NVSHMEM_HCA_LIST:-${hca_port_pairs_list}} export UCX_NET_DEVICES=${UCX_NET_DEVICES:-${hca_port_pairs_list}} echo "[Infer RoCE] NCCL_IB_HCA=${NCCL_IB_HCA}" echo "[Infer RoCE] NVSHMEM_HCA_LIST=${NVSHMEM_HCA_LIST}" echo "[Infer RoCE] UCX_NET_DEVICES=${UCX_NET_DEVICES}" else echo "[Infer RoCE] WARNING: No active RoCE HCAs found. NCCL_IB_HCA will not be set." fi if [ ${#active_hcas[@]} -gt 0 ]; then echo "[Infer RoCE] Finding GID_INDEX for each active HCA (SR-IOV compatible)..." # For SR-IOV environments, find the most common IPv4 RoCE v2 GID index across all HCAs declare -A gid_index_count declare -A hca_gid_index for hca_name in "${active_hcas[@]}"; do echo "[Infer RoCE] Processing HCA: ${hca_name}" # Find all RoCE v2 IPv4 GIDs for this HCA and count by index for tpath in /sys/class/infiniband/${hca_name}/ports/1/gid_attrs/types/*; do if grep -q "${KSERVE_INFER_IB_GID_INDEX_GREP}" "$tpath" 2>/dev/null; then idx=$(basename "$tpath") gid_file="/sys/class/infiniband/${hca_name}/ports/1/gids/${idx}" # Check for IPv4 GID (contains ffff:) if [ -f "$gid_file" ] && grep -q "ffff:" "$gid_file"; then gid_value=$(cat "$gid_file" 2>/dev/null || echo "") echo "[Infer RoCE] Found IPv4 RoCE v2 GID for ${hca_name}: index=${idx}, gid=${gid_value}" hca_gid_index["${hca_name}"]="${idx}" gid_index_count["${idx}"]=$((${gid_index_count["${idx}"]} + 1)) break # Use first found IPv4 GID per HCA fi fi done done # Find the most common GID index (most likely to be consistent across nodes) best_gid_index="" max_count=0 for idx in "${!gid_index_count[@]}"; do count=${gid_index_count["${idx}"]} echo "[Infer RoCE] GID_INDEX ${idx} found on ${count} HCAs" if [ $count -gt $max_count ]; then max_count=$count best_gid_index="$idx" fi done # Use deterministic fallback if tied - prefer index 3 (SR-IOV standard) if [ ${#gid_index_count[@]} -gt 1 ]; then echo "[Infer RoCE] Multiple GID indices found, selecting most common: ${best_gid_index}" # If there's a tie, prefer index 3 as it's most common in SR-IOV setups if [ -n "${gid_index_count['3']}" ] && [ "${gid_index_count['3']}" -eq "$max_count" ]; then best_gid_index="3" echo "[Infer RoCE] Using deterministic fallback: GID_INDEX=3 (SR-IOV standard)" fi fi # Check if GID_INDEX is already set via environment variables if [ -n "${NCCL_IB_GID_INDEX}" ]; then echo "[Infer RoCE] Using pre-configured NCCL_IB_GID_INDEX=${NCCL_IB_GID_INDEX} from environment" export NVSHMEM_IB_GID_INDEX=${NVSHMEM_IB_GID_INDEX:-$NCCL_IB_GID_INDEX} export UCX_IB_GID_INDEX=${UCX_IB_GID_INDEX:-$NCCL_IB_GID_INDEX} echo "[Infer RoCE] Using pre-configured GID_INDEX=${NCCL_IB_GID_INDEX} for NCCL, NVSHMEM, and UCX" elif [ -n "$best_gid_index" ]; then echo "[Infer RoCE] Selected GID_INDEX: ${best_gid_index} (found on ${max_count} HCAs)" export NCCL_IB_GID_INDEX=${NCCL_IB_GID_INDEX:-$best_gid_index} export NVSHMEM_IB_GID_INDEX=${NVSHMEM_IB_GID_INDEX:-$best_gid_index} export UCX_IB_GID_INDEX=${UCX_IB_GID_INDEX:-$best_gid_index} echo "[Infer RoCE] Exported GID_INDEX=${best_gid_index} for NCCL, NVSHMEM, and UCX" else echo "[Infer RoCE] ERROR: No valid IPv4 ${KSERVE_INFER_IB_GID_INDEX_GREP} GID_INDEX found on any HCA." fi else echo "[Infer RoCE] No active HCAs found, skipping GID_INDEX inference." fi fi # --disable-access-log-for-endpoints landed in vLLM 0.16.0 (vllm-project/vllm#30011). # Older versions still need the blanket --disable-uvicorn-access-log. ACCESS_LOG_ARGS="--disable-uvicorn-access-log" VLLM_VERSION=$(vllm --version 2>/dev/null | tail -1 | awk '{print $NF}') echo "[access-log-detect] vllm version='${VLLM_VERSION}'" if [[ "$VLLM_VERSION" =~ ^[0-9]+\.[0-9]+ ]] && [ "$(printf '%s\n%s\n' "0.16.0" "${VLLM_VERSION}" | sort -V | head -1)" = "0.16.0" ]; then ACCESS_LOG_ARGS="--disable-access-log-for-endpoints /health,/metrics,/ping" fi echo "[access-log-detect] selected ACCESS_LOG_ARGS='${ACCESS_LOG_ARGS}'" # --shutdown-timeout landed in vLLM 0.18.0 (vllm-project/vllm#36666). SHUTDOWN_TIMEOUT_ARGS="" if [[ "$VLLM_VERSION" =~ ^[0-9]+\.[0-9]+ ]] && [ "$(printf '%s\n%s\n' "0.18.0" "${VLLM_VERSION}" | sort -V | head -1)" = "0.18.0" ]; then SHUTDOWN_TIMEOUT_ARGS="--shutdown-timeout 40" fi eval "exec vllm serve /mnt/models \ --served-model-name "facebook/opt-125m" "publishers/kserve-ci-e2e-test/models/facebook/opt-125m" \ --port 8001 \ ${ACCESS_LOG_ARGS} \ ${SHUTDOWN_TIMEOUT_ARGS} \ --enable-ssl-refresh \ --ssl-certfile /var/run/kserve/tls/tls.crt \ --ssl-keyfile /var/run/kserve/tls/tls.key \ ${VLLM_ADDITIONAL_ARGS} \ $@" - -- env: - name: HOME value: /home - name: VLLM_LOGGING_LEVEL value: DEBUG - name: VLLM_CPU_KVCACHE_SPACE value: "1" - name: VLLM_ENABLE_V1_MULTIPROCESSING value: "0" - name: USER value: nonroot - name: TORCHINDUCTOR_CACHE_DIR value: /tmp/torchinductor-cache - name: HF_HUB_CACHE value: /models image: public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:v0.19.0 imagePullPolicy: IfNotPresent lifecycle: preStop: exec: command: - /bin/sleep - "15" livenessProbe: failureThreshold: 8 httpGet: path: /health port: 8000 scheme: HTTPS initialDelaySeconds: 180 periodSeconds: 30 successThreshold: 1 timeoutSeconds: 30 name: main ports: - containerPort: 8001 protocol: TCP readinessProbe: failureThreshold: 3 httpGet: path: /health port: 8000 scheme: HTTPS initialDelaySeconds: 30 periodSeconds: 10 successThreshold: 1 timeoutSeconds: 5 resources: limits: cpu: "2" memory: 7Gi requests: cpu: 200m memory: 2Gi securityContext: allowPrivilegeEscalation: false capabilities: drop: - ALL readOnlyRootFilesystem: true runAsNonRoot: true seccompProfile: type: RuntimeDefault startupProbe: failureThreshold: 60 httpGet: path: /health port: 8001 scheme: HTTPS periodSeconds: 10 successThreshold: 1 timeoutSeconds: 1 terminationMessagePath: /dev/termination-log terminationMessagePolicy: FallbackToLogsOnError volumeMounts: - mountPath: /home name: home - mountPath: /tmp name: tmp-dir - mountPath: /dev/shm name: dshm - mountPath: /models name: model-cache - mountPath: /var/run/kserve/tls name: tls-certs readOnly: true - mountPath: /mnt/models name: kserve-pvc-source readOnly: true dnsPolicy: ClusterFirst initContainers: - command: - /app/pd-sidecar - --port=8000 - --vllm-port=8001 - --kv-connector=nixlv2 - --enable-ssrf-protection=true - --pool-group=inference.networking.x-k8s.io - --inference-pool=kserve-ci-e2e-test/llmisvc-model-pvc-router-manage-e8706282-inference-pool - --secure-proxy=true - --cert-path=/var/run/kserve/tls - --decoder-use-tls=true - --prefiller-use-tls=true env: - name: INFERENCE_POOL_NAMESPACE valueFrom: fieldRef: apiVersion: v1 fieldPath: metadata.namespace - name: SSL_CERT_DIR value: /var/run/kserve/tls:/var/run/secrets/kubernetes.io/serviceaccount:/etc/pki/tls/certs - name: INFERENCE_POOL_NAME value: llmisvc-model-pvc-router-manage-e8706282-inference-pool image: ghcr.io/llm-d/llm-d-router-disagg-sidecar:v0.9.0-rc.2 imagePullPolicy: IfNotPresent livenessProbe: failureThreshold: 3 httpGet: path: /health port: 8000 scheme: HTTPS initialDelaySeconds: 10 periodSeconds: 10 successThreshold: 1 timeoutSeconds: 10 name: llm-d-routing-sidecar ports: - containerPort: 8000 protocol: TCP readinessProbe: failureThreshold: 10 httpGet: path: /health port: 8000 scheme: HTTPS initialDelaySeconds: 10 periodSeconds: 10 successThreshold: 1 timeoutSeconds: 5 resources: {} restartPolicy: Always securityContext: allowPrivilegeEscalation: false capabilities: drop: - ALL readOnlyRootFilesystem: true runAsNonRoot: true terminationMessagePath: /dev/termination-log terminationMessagePolicy: FallbackToLogsOnError volumeMounts: - mountPath: /var/run/kserve/tls name: tls-certs readOnly: true restartPolicy: Always schedulerName: default-scheduler securityContext: {} serviceAccount: llmisvc-model-pvc-router-manage-e8706282-kserve serviceAccountName: llmisvc-model-pvc-router-manage-e8706282-kserve terminationGracePeriodSeconds: 60 volumes: - emptyDir: {} name: home - emptyDir: {} name: tmp-dir - emptyDir: medium: Memory sizeLimit: 1Gi name: dshm - emptyDir: {} name: model-cache - name: tls-certs secret: defaultMode: 420 secretName: llmisvb19f98874e050eec8ca94d49676113f0-kserve-self-signed-certs - name: kserve-pvc-source persistentVolumeClaim: claimName: e2e-pvc-model-storage status: availableReplicas: 1 conditions: - lastTransitionTime: "2026-07-02T22:11:31Z" lastUpdateTime: "2026-07-02T22:11:31Z" message: Deployment has minimum availability. reason: MinimumReplicasAvailable status: "True" type: Available - lastTransitionTime: "2026-07-02T22:08:28Z" lastUpdateTime: "2026-07-02T22:11:31Z" message: ReplicaSet "llmisvc-model-pvc-router-manage-e8706282-kserve-7db79954c" has successfully progressed. reason: NewReplicaSetAvailable status: "True" type: Progressing observedGeneration: 1 readyReplicas: 1 replicas: 1 updatedReplicas: 1 - apiVersion: apps/v1 kind: Deployment metadata: annotations: deployment.kubernetes.io/revision: "1" creationTimestamp: "2026-07-02T22:08:28Z" generation: 1 labels: app.kubernetes.io/component: llminferenceservice-workload-prefill app.kubernetes.io/name: llmisvc-model-pvc-router-manage-e8706282 app.kubernetes.io/part-of: llminferenceservice kserve.io/component: workload llm-d.ai/role: prefill 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":"9f190595-570c-42e2-b950-44b09c28e1f5"}: {} 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:command: {} f:env: .: {} k:{"name":"HF_HUB_CACHE"}: .: {} f:name: {} f:value: {} k:{"name":"HOME"}: .: {} f:name: {} f:value: {} k:{"name":"TORCHINDUCTOR_CACHE_DIR"}: .: {} f:name: {} f:value: {} k:{"name":"USER"}: .: {} f:name: {} f:value: {} k:{"name":"VLLM_CPU_KVCACHE_SPACE"}: .: {} f:name: {} f:value: {} k:{"name":"VLLM_ENABLE_V1_MULTIPROCESSING"}: .: {} f:name: {} f:value: {} 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:initialDelaySeconds: {} f:periodSeconds: {} f:successThreshold: {} f:timeoutSeconds: {} f:name: {} f:ports: .: {} k:{"containerPort":8000,"protocol":"TCP"}: .: {} f:containerPort: {} f:protocol: {} f:readinessProbe: .: {} f:failureThreshold: {} f:httpGet: .: {} f:path: {} f:port: {} f:scheme: {} 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: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":"/mnt/models"}: .: {} f:mountPath: {} f:name: {} f:readOnly: {} 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":"kserve-pvc-source"}: .: {} f:name: {} f:persistentVolumeClaim: .: {} f:claimName: {} 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: "2026-07-02T22:08:28Z" - 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:updatedReplicas: {} manager: kube-controller-manager operation: Update subresource: status time: "2026-07-02T22:11:31Z" name: llmisvc-model-pvc-router-manage-e8706282-kserve-prefill namespace: kserve-ci-e2e-test ownerReferences: - apiVersion: serving.kserve.io/v1alpha2 blockOwnerDeletion: true controller: true kind: LLMInferenceService name: llmisvc-model-pvc-router-manage-e8706282 uid: 9f190595-570c-42e2-b950-44b09c28e1f5 resourceVersion: "60822" uid: 09722021-ed72-4dc4-aa0c-aab47e0b4fe9 spec: progressDeadlineSeconds: 600 replicas: 1 revisionHistoryLimit: 10 selector: matchLabels: app.kubernetes.io/component: llminferenceservice-workload-prefill app.kubernetes.io/name: llmisvc-model-pvc-router-manage-e8706282 app.kubernetes.io/part-of: llminferenceservice kserve.io/component: workload llm-d.ai/role: prefill strategy: rollingUpdate: maxSurge: 25% maxUnavailable: 25% type: RollingUpdate template: metadata: labels: app.kubernetes.io/component: llminferenceservice-workload-prefill app.kubernetes.io/name: llmisvc-model-pvc-router-manage-e8706282 app.kubernetes.io/part-of: llminferenceservice kserve.io/component: workload llm-d.ai/role: prefill spec: containers: - command: - /bin/bash - -c - |- if [ -f /etc/profile.d/ibm-aiu-setup.sh ]; then source /etc/profile.d/ibm-aiu-setup.sh fi if [ "$KSERVE_INFER_ROCE" = "true" ]; then echo "Trying to infer RoCE configs ... " grep -H . /sys/class/infiniband/*/ports/*/gids/* 2>/dev/null grep -H . /sys/class/infiniband/*/ports/*/gid_attrs/types/* 2>/dev/null cat /proc/driver/nvidia/params KSERVE_INFER_IB_GID_INDEX_GREP=${KSERVE_INFER_IB_GID_INDEX_GREP:-"RoCE v2"} echo "[Infer RoCE] Discovering active HCAs ..." active_hcas=() # Loop through all mlx5 devices found in sysfs for hca_dir in /sys/class/infiniband/mlx5_*; do # Ensure it's a directory before proceeding if [ -d "$hca_dir" ]; then hca_name=$(basename "$hca_dir") port_state_file="$hca_dir/ports/1/state" # Assume port 1 type_file="$hca_dir/ports/1/gid_attrs/types/*" echo "[Infer RoCE] Check if the port state file ${port_state_file} exists and contains 'ACTIVE'" if [ -f "$port_state_file" ] && grep -q "ACTIVE" "$port_state_file" && grep -q "${KSERVE_INFER_IB_GID_INDEX_GREP}" ${type_file} 2>/dev/null; then echo "[Infer RoCE] Found active HCA: $hca_name" active_hcas+=("$hca_name") else echo "[Infer RoCE] Skipping inactive or down HCA: $hca_name" fi fi done # Check if we found any active HCAs if [ ${#active_hcas[@]} -gt 0 ]; then # Join the array elements with a comma hca_port_pairs=() for hca in "${active_hcas[@]}"; do hca_port_pairs+=("${hca}:1") done active_hca_list=$(IFS=,; echo "${active_hcas[*]}") hca_port_pairs_list=$(IFS=,; echo "${hca_port_pairs[*]}") echo "[Infer RoCE] Setting active HCAs: ${active_hca_list}" export NCCL_IB_HCA=${NCCL_IB_HCA:-${active_hca_list}} export NVSHMEM_HCA_LIST=${NVSHMEM_HCA_LIST:-${hca_port_pairs_list}} export UCX_NET_DEVICES=${UCX_NET_DEVICES:-${hca_port_pairs_list}} echo "[Infer RoCE] NCCL_IB_HCA=${NCCL_IB_HCA}" echo "[Infer RoCE] NVSHMEM_HCA_LIST=${NVSHMEM_HCA_LIST}" echo "[Infer RoCE] UCX_NET_DEVICES=${UCX_NET_DEVICES}" else echo "[Infer RoCE] WARNING: No active RoCE HCAs found. NCCL_IB_HCA will not be set." fi if [ ${#active_hcas[@]} -gt 0 ]; then echo "[Infer RoCE] Finding GID_INDEX for each active HCA (SR-IOV compatible)..." # For SR-IOV environments, find the most common IPv4 RoCE v2 GID index across all HCAs declare -A gid_index_count declare -A hca_gid_index for hca_name in "${active_hcas[@]}"; do echo "[Infer RoCE] Processing HCA: ${hca_name}" # Find all RoCE v2 IPv4 GIDs for this HCA and count by index for tpath in /sys/class/infiniband/${hca_name}/ports/1/gid_attrs/types/*; do if grep -q "${KSERVE_INFER_IB_GID_INDEX_GREP}" "$tpath" 2>/dev/null; then idx=$(basename "$tpath") gid_file="/sys/class/infiniband/${hca_name}/ports/1/gids/${idx}" # Check for IPv4 GID (contains ffff:) if [ -f "$gid_file" ] && grep -q "ffff:" "$gid_file"; then gid_value=$(cat "$gid_file" 2>/dev/null || echo "") echo "[Infer RoCE] Found IPv4 RoCE v2 GID for ${hca_name}: index=${idx}, gid=${gid_value}" hca_gid_index["${hca_name}"]="${idx}" gid_index_count["${idx}"]=$((${gid_index_count["${idx}"]} + 1)) break # Use first found IPv4 GID per HCA fi fi done done # Find the most common GID index (most likely to be consistent across nodes) best_gid_index="" max_count=0 for idx in "${!gid_index_count[@]}"; do count=${gid_index_count["${idx}"]} echo "[Infer RoCE] GID_INDEX ${idx} found on ${count} HCAs" if [ $count -gt $max_count ]; then max_count=$count best_gid_index="$idx" fi done # Use deterministic fallback if tied - prefer index 3 (SR-IOV standard) if [ ${#gid_index_count[@]} -gt 1 ]; then echo "[Infer RoCE] Multiple GID indices found, selecting most common: ${best_gid_index}" # If there's a tie, prefer index 3 as it's most common in SR-IOV setups if [ -n "${gid_index_count['3']}" ] && [ "${gid_index_count['3']}" -eq "$max_count" ]; then best_gid_index="3" echo "[Infer RoCE] Using deterministic fallback: GID_INDEX=3 (SR-IOV standard)" fi fi # Check if GID_INDEX is already set via environment variables if [ -n "${NCCL_IB_GID_INDEX}" ]; then echo "[Infer RoCE] Using pre-configured NCCL_IB_GID_INDEX=${NCCL_IB_GID_INDEX} from environment" export NVSHMEM_IB_GID_INDEX=${NVSHMEM_IB_GID_INDEX:-$NCCL_IB_GID_INDEX} export UCX_IB_GID_INDEX=${UCX_IB_GID_INDEX:-$NCCL_IB_GID_INDEX} echo "[Infer RoCE] Using pre-configured GID_INDEX=${NCCL_IB_GID_INDEX} for NCCL, NVSHMEM, and UCX" elif [ -n "$best_gid_index" ]; then echo "[Infer RoCE] Selected GID_INDEX: ${best_gid_index} (found on ${max_count} HCAs)" export NCCL_IB_GID_INDEX=${NCCL_IB_GID_INDEX:-$best_gid_index} export NVSHMEM_IB_GID_INDEX=${NVSHMEM_IB_GID_INDEX:-$best_gid_index} export UCX_IB_GID_INDEX=${UCX_IB_GID_INDEX:-$best_gid_index} echo "[Infer RoCE] Exported GID_INDEX=${best_gid_index} for NCCL, NVSHMEM, and UCX" else echo "[Infer RoCE] ERROR: No valid IPv4 ${KSERVE_INFER_IB_GID_INDEX_GREP} GID_INDEX found on any HCA." fi else echo "[Infer RoCE] No active HCAs found, skipping GID_INDEX inference." fi fi # --disable-access-log-for-endpoints landed in vLLM 0.16.0 (vllm-project/vllm#30011). # Older versions still need the blanket --disable-uvicorn-access-log. ACCESS_LOG_ARGS="--disable-uvicorn-access-log" VLLM_VERSION=$(vllm --version 2>/dev/null | tail -1 | awk '{print $NF}') echo "[access-log-detect] vllm version='${VLLM_VERSION}'" if [[ "$VLLM_VERSION" =~ ^[0-9]+\.[0-9]+ ]] && [ "$(printf '%s\n%s\n' "0.16.0" "${VLLM_VERSION}" | sort -V | head -1)" = "0.16.0" ]; then ACCESS_LOG_ARGS="--disable-access-log-for-endpoints /health,/metrics,/ping" fi echo "[access-log-detect] selected ACCESS_LOG_ARGS='${ACCESS_LOG_ARGS}'" # --shutdown-timeout landed in vLLM 0.18.0 (vllm-project/vllm#36666). SHUTDOWN_TIMEOUT_ARGS="" if [[ "$VLLM_VERSION" =~ ^[0-9]+\.[0-9]+ ]] && [ "$(printf '%s\n%s\n' "0.18.0" "${VLLM_VERSION}" | sort -V | head -1)" = "0.18.0" ]; then SHUTDOWN_TIMEOUT_ARGS="--shutdown-timeout 40" fi eval "exec vllm serve /mnt/models \ --served-model-name "facebook/opt-125m" \ --port 8000 \ ${ACCESS_LOG_ARGS} \ ${SHUTDOWN_TIMEOUT_ARGS} \ --enable-ssl-refresh \ --ssl-certfile /var/run/kserve/tls/tls.crt \ --ssl-keyfile /var/run/kserve/tls/tls.key \ ${VLLM_ADDITIONAL_ARGS} \ $@" - -- env: - name: HOME value: /home - name: VLLM_LOGGING_LEVEL value: DEBUG - name: VLLM_CPU_KVCACHE_SPACE value: "1" - name: VLLM_ENABLE_V1_MULTIPROCESSING value: "0" - name: USER value: nonroot - name: TORCHINDUCTOR_CACHE_DIR value: /tmp/torchinductor-cache - name: HF_HUB_CACHE value: /models image: public.ecr.aws/q9t5s3a7/vllm-cpu-release-repo:v0.19.0 imagePullPolicy: IfNotPresent lifecycle: preStop: exec: command: - /bin/sleep - "15" livenessProbe: failureThreshold: 8 httpGet: path: /health port: 8000 scheme: HTTPS initialDelaySeconds: 180 periodSeconds: 30 successThreshold: 1 timeoutSeconds: 30 name: main ports: - containerPort: 8000 protocol: TCP readinessProbe: failureThreshold: 3 httpGet: path: /health port: 8000 scheme: HTTPS initialDelaySeconds: 30 periodSeconds: 10 successThreshold: 1 timeoutSeconds: 5 resources: limits: cpu: "2" memory: 7Gi requests: cpu: 200m memory: 2Gi securityContext: allowPrivilegeEscalation: false capabilities: drop: - ALL readOnlyRootFilesystem: true 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: File volumeMounts: - mountPath: /home name: home - mountPath: /tmp name: tmp-dir - mountPath: /dev/shm name: dshm - mountPath: /models name: model-cache - mountPath: /var/run/kserve/tls name: tls-certs readOnly: true - mountPath: /mnt/models name: kserve-pvc-source readOnly: true dnsPolicy: ClusterFirst restartPolicy: Always schedulerName: default-scheduler securityContext: {} terminationGracePeriodSeconds: 60 volumes: - emptyDir: {} name: home - emptyDir: {} name: tmp-dir - emptyDir: medium: Memory sizeLimit: 1Gi name: dshm - emptyDir: {} name: model-cache - name: tls-certs secret: defaultMode: 420 secretName: llmisvb19f98874e050eec8ca94d49676113f0-kserve-self-signed-certs - name: kserve-pvc-source persistentVolumeClaim: claimName: e2e-pvc-model-storage status: availableReplicas: 1 conditions: - lastTransitionTime: "2026-07-02T22:11:31Z" lastUpdateTime: "2026-07-02T22:11:31Z" message: Deployment has minimum availability. reason: MinimumReplicasAvailable status: "True" type: Available - lastTransitionTime: "2026-07-02T22:08:28Z" lastUpdateTime: "2026-07-02T22:11:31Z" message: ReplicaSet "llmisvc-model-pvc-router-manage-e8706282-kserve-prefill-79497db4cc" has successfully progressed. reason: NewReplicaSetAvailable status: "True" type: Progressing observedGeneration: 1 readyReplicas: 1 replicas: 1 updatedReplicas: 1 - apiVersion: apps/v1 kind: Deployment metadata: annotations: deployment.kubernetes.io/revision: "1" creationTimestamp: "2026-07-02T22:10:51Z" generation: 1 labels: app.kubernetes.io/component: llminferenceservice-router-scheduler app.kubernetes.io/name: llmisvc-model-fb-opt-125m-route-50bc673d app.kubernetes.io/part-of: llminferenceservice 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:ownerReferences: .: {} k:{"uid":"7e453f7e-1de8-49d5-a8fa-024a9229a616"}: {} f:spec: f:progressDeadlineSeconds: {} f:replicas: {} f:revisionHistoryLimit: {} f:selector: {} f:strategy: f:type: {} f:template: f:metadata: f:annotations: .: {} f:app.kubernetes.io/version: {} f:certificates.kserve.io/expiration-v2: {} f:labels: .: {} f:app.kubernetes.io/component: {} f:app.kubernetes.io/name: {} f:app.kubernetes.io/part-of: {} f:spec: f:containers: k:{"name":"main"}: .: {} f:args: {} 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: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":"/tmp/tokenizer"}: .: {} f:mountPath: {} f:name: {} k:{"mountPath":"/var/run/kserve/tls"}: .: {} f:mountPath: {} f:name: {} f:readOnly: {} f:dnsPolicy: {} f:initContainers: .: {} k:{"name":"storage-initializer"}: .: {} f:args: {} f:env: .: {} k:{"name":"AWS_ACCESS_KEY_ID"}: .: {} f:name: {} f:valueFrom: .: {} f:secretKeyRef: {} k:{"name":"AWS_CA_BUNDLE"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_CA_BUNDLE_CONFIGMAP"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_ENDPOINT_URL"}: .: {} f:name: {} f:value: {} k:{"name":"AWS_SECRET_ACCESS_KEY"}: .: {} f:name: {} f:valueFrom: .: {} f:secretKeyRef: {} k:{"name":"HF_HUB_ENABLE_HF_TRANSFER"}: .: {} f:name: {} f:value: {} k:{"name":"HF_XET_HIGH_PERFORMANCE"}: .: {} f:name: {} f:value: {} k:{"name":"HF_XET_NUM_CONCURRENT_RANGE_GETS"}: .: {} f:name: {} f:value: {} k:{"name":"S3_ENDPOINT"}: .: {} f:name: {} f:value: {} k:{"name":"S3_USE_HTTPS"}: .: {} f:name: {} f:value: {} k:{"name":"S3_VERIFY_SSL"}: .: {} f:name: {} f:value: {} k:{"name":"STORAGE_ALLOW_PATTERNS"}: .: {} f:name: {} f:value: {} f:image: {} f:imagePullPolicy: {} f:name: {} f:resources: .: {} f:limits: .: {} f:cpu: {} f:memory: {} f:requests: .: {} f:cpu: {} f:memory: {} f:terminationMessagePath: {} f:terminationMessagePolicy: {} f:volumeMounts: .: {} k:{"mountPath":"/mnt/models"}: .: {} f:mountPath: {} f:name: {} f:restartPolicy: {} f:schedulerName: {} f:securityContext: {} f:serviceAccount: {} f:serviceAccountName: {} f:terminationGracePeriodSeconds: {} f:volumes: .: {} k:{"name":"kserve-provision-location"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"tls-certs"}: .: {} f:name: {} f:secret: .: {} f:defaultMode: {} f:secretName: {} k:{"name":"tokenizer-cache"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"tokenizer-tmp"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"tokenizer-uds"}: .: {} f:emptyDir: {} f:name: {} manager: manager operation: Update time: "2026-07-02T22:10:51Z" - 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:updatedReplicas: {} manager: kube-controller-manager operation: Update subresource: status time: "2026-07-02T22:11:25Z" name: llmisvc44d181485fad85e662eb092f3749502f-kserve-router-scheduler namespace: kserve-ci-e2e-test ownerReferences: - apiVersion: serving.kserve.io/v1alpha2 blockOwnerDeletion: true controller: true kind: LLMInferenceService name: llmisvc-model-fb-opt-125m-route-50bc673d uid: 7e453f7e-1de8-49d5-a8fa-024a9229a616 resourceVersion: "60719" uid: 6d001824-e41f-4e8b-9ebe-78a483131dc5 spec: progressDeadlineSeconds: 600 replicas: 1 revisionHistoryLimit: 10 selector: matchLabels: app.kubernetes.io/component: llminferenceservice-router-scheduler app.kubernetes.io/name: llmisvc-model-fb-opt-125m-route-50bc673d app.kubernetes.io/part-of: llminferenceservice strategy: type: Recreate template: metadata: annotations: app.kubernetes.io/version: 0.9.0 certificates.kserve.io/expiration-v2: "true" labels: app.kubernetes.io/component: llminferenceservice-router-scheduler app.kubernetes.io/name: llmisvc-model-fb-opt-125m-route-50bc673d app.kubernetes.io/part-of: llminferenceservice spec: containers: - args: - --config-text - | apiVersion: inference.networking.x-k8s.io/v1alpha1 kind: EndpointPickerConfig plugins: - type: disagg-headers-handler - type: prefill-filter - type: decode-filter - type: queue-scorer - type: prefix-cache-scorer - type: max-score-picker - type: always-disagg-pd-decider - parameters: deciders: prefill: always-disagg-pd-decider type: disagg-profile-handler schedulingProfiles: - name: prefill plugins: - pluginRef: prefill-filter - pluginRef: queue-scorer weight: 2 - pluginRef: prefix-cache-scorer weight: 3 - pluginRef: max-score-picker - name: decode plugins: - pluginRef: decode-filter - pluginRef: queue-scorer weight: 2 - pluginRef: prefix-cache-scorer weight: 3 - pluginRef: max-score-picker command: - /app/epp - --pool-name - llmisvc-model-fb-opt-125m-route-50bc673d-inference-pool - --pool-namespace - kserve-ci-e2e-test - --zap-encoder - json - --grpc-port - "9002" - --grpc-health-port - "9003" - --enable-cert-reload=true - --secure-serving=true - --model-server-metrics-scheme=https - --cert-path=/var/run/kserve/tls env: - name: SSL_CERT_DIR value: /var/run/kserve/tls:/var/run/secrets/kubernetes.io/serviceaccount:/etc/pki/tls/certs image: ghcr.io/llm-d/llm-d-router-endpoint-picker:v0.9.0-rc.2 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: requests: cpu: 256m memory: 500Mi 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 - mountPath: /tmp/tokenizer name: tokenizer-uds dnsPolicy: ClusterFirst initContainers: - args: - hf://facebook/opt-125m - /mnt/models env: - name: AWS_ACCESS_KEY_ID valueFrom: secretKeyRef: key: AWS_ACCESS_KEY_ID name: seaweedfs-s3-creds - name: AWS_SECRET_ACCESS_KEY valueFrom: secretKeyRef: key: AWS_SECRET_ACCESS_KEY name: seaweedfs-s3-creds - name: S3_USE_HTTPS value: "0" - name: S3_ENDPOINT value: s3-service.kserve:8333 - name: AWS_ENDPOINT_URL value: http://s3-service.kserve:8333 - name: S3_VERIFY_SSL value: "0" - name: AWS_CA_BUNDLE value: /etc/ssl/custom-certs/cabundle.crt - name: AWS_CA_BUNDLE_CONFIGMAP value: odh-kserve-custom-ca-bundle - name: HF_HUB_ENABLE_HF_TRANSFER value: "1" - name: HF_XET_HIGH_PERFORMANCE value: "1" - name: HF_XET_NUM_CONCURRENT_RANGE_GETS value: "8" - name: STORAGE_ALLOW_PATTERNS value: '["tokenizer.json", "tokenizer_config.json", "special_tokens_map.json", "vocab.json", "merges.txt", "config.json", "generation_config.json"]' image: quay.io/opendatahub/kserve-storage-initializer@sha256:33f82c067d520fe749f73749cf76b5935fe69bd9eb141504c6660698c7947743 imagePullPolicy: IfNotPresent name: storage-initializer resources: limits: cpu: "1" memory: 24Gi requests: cpu: 100m memory: 100Mi terminationMessagePath: /dev/termination-log terminationMessagePolicy: FallbackToLogsOnError volumeMounts: - mountPath: /mnt/models name: kserve-provision-location restartPolicy: Always schedulerName: default-scheduler securityContext: {} serviceAccount: llmisvc-model-fb-opt-125m-route-50bc673d-epp-sa serviceAccountName: llmisvc-model-fb-opt-125m-route-50bc673d-epp-sa terminationGracePeriodSeconds: 60 volumes: - name: tls-certs secret: defaultMode: 420 secretName: llmisv44d181485fad85e662eb092f3749502f-kserve-self-signed-certs - emptyDir: {} name: tokenizer-uds - emptyDir: {} name: tokenizer-tmp - emptyDir: {} name: tokenizer-cache - emptyDir: {} name: kserve-provision-location status: availableReplicas: 1 conditions: - lastTransitionTime: "2026-07-02T22:11:25Z" lastUpdateTime: "2026-07-02T22:11:25Z" message: Deployment has minimum availability. reason: MinimumReplicasAvailable status: "True" type: Available - lastTransitionTime: "2026-07-02T22:10:51Z" lastUpdateTime: "2026-07-02T22:11:25Z" message: ReplicaSet "llmisvc44d181485fad85e662eb092f3749502f-kserve-router-scheduler-69454684c7" has successfully progressed. reason: NewReplicaSetAvailable status: "True" type: Progressing observedGeneration: 1 readyReplicas: 1 replicas: 1 updatedReplicas: 1 - apiVersion: apps/v1 kind: Deployment metadata: annotations: deployment.kubernetes.io/revision: "1" creationTimestamp: "2026-07-02T22:08:29Z" generation: 1 labels: app.kubernetes.io/component: llminferenceservice-router-scheduler app.kubernetes.io/name: llmisvc-model-pvc-router-manage-e8706282 app.kubernetes.io/part-of: llminferenceservice 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:ownerReferences: .: {} k:{"uid":"9f190595-570c-42e2-b950-44b09c28e1f5"}: {} f:spec: f:progressDeadlineSeconds: {} f:replicas: {} f:revisionHistoryLimit: {} f:selector: {} f:strategy: f:type: {} f:template: f:metadata: f:annotations: .: {} f:app.kubernetes.io/version: {} f:certificates.kserve.io/expiration-v2: {} f:labels: .: {} f:app.kubernetes.io/component: {} f:app.kubernetes.io/name: {} f:app.kubernetes.io/part-of: {} f:spec: f:containers: k:{"name":"main"}: .: {} f:args: {} 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: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":"/tmp/tokenizer"}: .: {} 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:serviceAccount: {} f:serviceAccountName: {} f:terminationGracePeriodSeconds: {} f:volumes: .: {} k:{"name":"kserve-pvc-source"}: .: {} f:name: {} f:persistentVolumeClaim: .: {} f:claimName: {} k:{"name":"tls-certs"}: .: {} f:name: {} f:secret: .: {} f:defaultMode: {} f:secretName: {} k:{"name":"tokenizer-cache"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"tokenizer-tmp"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"tokenizer-uds"}: .: {} f:emptyDir: {} f:name: {} manager: manager operation: Update time: "2026-07-02T22:08:29Z" - apiVersion: apps/v1 fieldsType: FieldsV1 fieldsV1: f:metadata: f:annotations: .: {} f:deployment.kubernetes.io/revision: {} f:status: f:availableReplicas: {} f:conditions: .: {} k:{"type":"Available"}: .: {} f:lastTransitionTime: {} f:lastUpdateTime: {} f:message: {} f:reason: {} f:status: {} f:type: {} k:{"type":"Progressing"}: .: {} f:lastTransitionTime: {} f:lastUpdateTime: {} f:message: {} f:reason: {} f:status: {} f:type: {} f:observedGeneration: {} f:readyReplicas: {} f:replicas: {} f:updatedReplicas: {} manager: kube-controller-manager operation: Update subresource: status time: "2026-07-02T22:09:01Z" name: llmisvcb19f98874e050eec8ca94d49676113f0-kserve-router-scheduler namespace: kserve-ci-e2e-test ownerReferences: - apiVersion: serving.kserve.io/v1alpha2 blockOwnerDeletion: true controller: true kind: LLMInferenceService name: llmisvc-model-pvc-router-manage-e8706282 uid: 9f190595-570c-42e2-b950-44b09c28e1f5 resourceVersion: "58379" uid: ed01f41b-097b-485d-9c4e-392a28f96ef6 spec: progressDeadlineSeconds: 600 replicas: 1 revisionHistoryLimit: 10 selector: matchLabels: app.kubernetes.io/component: llminferenceservice-router-scheduler app.kubernetes.io/name: llmisvc-model-pvc-router-manage-e8706282 app.kubernetes.io/part-of: llminferenceservice strategy: type: Recreate template: metadata: annotations: app.kubernetes.io/version: 0.9.0 certificates.kserve.io/expiration-v2: "true" labels: app.kubernetes.io/component: llminferenceservice-router-scheduler app.kubernetes.io/name: llmisvc-model-pvc-router-manage-e8706282 app.kubernetes.io/part-of: llminferenceservice spec: containers: - args: - --config-text - | apiVersion: inference.networking.x-k8s.io/v1alpha1 kind: EndpointPickerConfig plugins: - type: disagg-headers-handler - type: prefill-filter - type: decode-filter - type: queue-scorer - type: prefix-cache-scorer - type: max-score-picker - type: always-disagg-pd-decider - parameters: deciders: prefill: always-disagg-pd-decider type: disagg-profile-handler schedulingProfiles: - name: prefill plugins: - pluginRef: prefill-filter - pluginRef: queue-scorer weight: 2 - pluginRef: prefix-cache-scorer weight: 3 - pluginRef: max-score-picker - name: decode plugins: - pluginRef: decode-filter - pluginRef: queue-scorer weight: 2 - pluginRef: prefix-cache-scorer weight: 3 - pluginRef: max-score-picker command: - /app/epp - --pool-name - llmisvc-model-pvc-router-manage-e8706282-inference-pool - --pool-namespace - kserve-ci-e2e-test - --zap-encoder - json - --grpc-port - "9002" - --grpc-health-port - "9003" - --enable-cert-reload=true - --secure-serving=true - --model-server-metrics-scheme=https - --cert-path=/var/run/kserve/tls env: - name: SSL_CERT_DIR value: /var/run/kserve/tls:/var/run/secrets/kubernetes.io/serviceaccount:/etc/pki/tls/certs image: ghcr.io/llm-d/llm-d-router-endpoint-picker:v0.9.0-rc.2 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: requests: cpu: 256m memory: 500Mi 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 - mountPath: /tmp/tokenizer name: tokenizer-uds dnsPolicy: ClusterFirst restartPolicy: Always schedulerName: default-scheduler securityContext: {} serviceAccount: llmisvc-model-pvc-router-manage-e8706282-epp-sa serviceAccountName: llmisvc-model-pvc-router-manage-e8706282-epp-sa terminationGracePeriodSeconds: 60 volumes: - name: tls-certs secret: defaultMode: 420 secretName: llmisvb19f98874e050eec8ca94d49676113f0-kserve-self-signed-certs - emptyDir: {} name: tokenizer-uds - emptyDir: {} name: tokenizer-tmp - emptyDir: {} name: tokenizer-cache - name: kserve-pvc-source persistentVolumeClaim: claimName: e2e-pvc-model-storage status: availableReplicas: 1 conditions: - lastTransitionTime: "2026-07-02T22:09:01Z" lastUpdateTime: "2026-07-02T22:09:01Z" message: Deployment has minimum availability. reason: MinimumReplicasAvailable status: "True" type: Available - lastTransitionTime: "2026-07-02T22:08:29Z" lastUpdateTime: "2026-07-02T22:09:01Z" message: ReplicaSet "llmisvcb19f98874e050eec8ca94d49676113f0-kserve-router-scheduler-67b4bb9646" has successfully progressed. reason: NewReplicaSetAvailable status: "True" type: Progressing observedGeneration: 1 readyReplicas: 1 replicas: 1 updatedReplicas: 1 - apiVersion: apps/v1 kind: Deployment metadata: annotations: deployment.kubernetes.io/revision: "1" creationTimestamp: "2026-07-02T21:31:26Z" generation: 1 labels: gateway.istio.io/managed: openshift.io-gateway-controller-v1 gateway.networking.k8s.io/gateway-name: router-gateway-1 managedFields: - apiVersion: apps/v1 fieldsType: FieldsV1 fieldsV1: f:metadata: f:annotations: {} f:labels: f:gateway.istio.io/managed: {} f:gateway.networking.k8s.io/gateway-name: {} f:ownerReferences: k:{"uid":"5292cc85-fb77-4560-a038-f3b40e91323c"}: {} f:spec: f:selector: {} f:template: f:metadata: f:annotations: f:istio.io/rev: {} f:prometheus.io/path: {} f:prometheus.io/port: {} f:prometheus.io/scrape: {} f:labels: f:gateway.istio.io/managed: {} f:gateway.networking.k8s.io/gateway-name: {} f:service.istio.io/canonical-name: {} f:service.istio.io/canonical-revision: {} f:sidecar.istio.io/inject: {} f:spec: f:containers: k:{"name":"istio-proxy"}: .: {} f:args: {} f:env: k:{"name":"CA_ADDR"}: .: {} f:name: {} f:value: {} k:{"name":"GOMAXPROCS"}: .: {} f:name: {} f:valueFrom: f:resourceFieldRef: {} k:{"name":"GOMEMLIMIT"}: .: {} f:name: {} f:valueFrom: f:resourceFieldRef: {} k:{"name":"HOST_IP"}: .: {} f:name: {} f:valueFrom: f:fieldRef: {} k:{"name":"INSTANCE_IP"}: .: {} f:name: {} f:valueFrom: f:fieldRef: {} k:{"name":"ISTIO_CPU_LIMIT"}: .: {} f:name: {} f:valueFrom: f:resourceFieldRef: {} k:{"name":"ISTIO_META_APP_CONTAINERS"}: .: {} f:name: {} f:value: {} k:{"name":"ISTIO_META_CLUSTER_ID"}: .: {} f:name: {} f:value: {} k:{"name":"ISTIO_META_INTERCEPTION_MODE"}: .: {} f:name: {} f:value: {} k:{"name":"ISTIO_META_MESH_ID"}: .: {} f:name: {} f:value: {} k:{"name":"ISTIO_META_NODE_NAME"}: .: {} f:name: {} f:valueFrom: f:fieldRef: {} k:{"name":"ISTIO_META_OWNER"}: .: {} f:name: {} f:value: {} k:{"name":"ISTIO_META_POD_PORTS"}: .: {} f:name: {} f:value: {} k:{"name":"ISTIO_META_WORKLOAD_NAME"}: .: {} f:name: {} f:value: {} k:{"name":"PILOT_CERT_PROVIDER"}: .: {} f:name: {} f:value: {} k:{"name":"POD_NAME"}: .: {} f:name: {} f:valueFrom: f:fieldRef: {} k:{"name":"POD_NAMESPACE"}: .: {} f:name: {} f:valueFrom: f:fieldRef: {} k:{"name":"PROXY_CONFIG"}: .: {} f:name: {} f:value: {} k:{"name":"SERVICE_ACCOUNT"}: .: {} f:name: {} f:valueFrom: f:fieldRef: {} k:{"name":"TRUST_DOMAIN"}: .: {} f:name: {} f:value: {} f:image: {} f:name: {} f:ports: k:{"containerPort":15020,"protocol":"TCP"}: .: {} f:containerPort: {} f:name: {} f:protocol: {} k:{"containerPort":15021,"protocol":"TCP"}: .: {} f:containerPort: {} f:name: {} f:protocol: {} k:{"containerPort":15090,"protocol":"TCP"}: .: {} f:containerPort: {} f:name: {} f:protocol: {} f:readinessProbe: f:failureThreshold: {} f:httpGet: f:path: {} f:port: {} f:scheme: {} 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:privileged: {} f:readOnlyRootFilesystem: {} f:runAsGroup: {} f:runAsNonRoot: {} f:runAsUser: {} f:startupProbe: f:failureThreshold: {} f:httpGet: f:path: {} f:port: {} f:scheme: {} f:initialDelaySeconds: {} f:periodSeconds: {} f:successThreshold: {} f:timeoutSeconds: {} f:volumeMounts: k:{"mountPath":"/etc/istio/pod"}: .: {} f:mountPath: {} f:name: {} k:{"mountPath":"/etc/istio/proxy"}: .: {} f:mountPath: {} f:name: {} k:{"mountPath":"/var/lib/istio/data"}: .: {} f:mountPath: {} f:name: {} k:{"mountPath":"/var/run/secrets/credential-uds"}: .: {} f:mountPath: {} f:name: {} k:{"mountPath":"/var/run/secrets/istio"}: .: {} f:mountPath: {} f:name: {} k:{"mountPath":"/var/run/secrets/tokens"}: .: {} f:mountPath: {} f:name: {} k:{"mountPath":"/var/run/secrets/workload-spiffe-credentials"}: .: {} f:mountPath: {} f:name: {} k:{"mountPath":"/var/run/secrets/workload-spiffe-uds"}: .: {} f:mountPath: {} f:name: {} f:securityContext: f:sysctls: {} f:serviceAccountName: {} f:volumes: k:{"name":"credential-socket"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"istio-data"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"istio-envoy"}: .: {} f:emptyDir: f:medium: {} f:name: {} k:{"name":"istio-podinfo"}: .: {} f:downwardAPI: f:items: {} f:name: {} k:{"name":"istio-token"}: .: {} f:name: {} f:projected: f:sources: {} k:{"name":"istiod-ca-cert"}: .: {} f:configMap: f:name: {} f:name: {} k:{"name":"workload-certs"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"workload-socket"}: .: {} f:emptyDir: {} f:name: {} manager: openshift.io/gateway-controller/v1 operation: Apply time: "2026-07-02T21:31:26Z" - 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:updatedReplicas: {} manager: kube-controller-manager operation: Update subresource: status time: "2026-07-02T21:31:51Z" name: router-gateway-1-openshift-default namespace: kserve-ci-e2e-test ownerReferences: - apiVersion: gateway.networking.k8s.io/v1beta1 kind: Gateway name: router-gateway-1 uid: 5292cc85-fb77-4560-a038-f3b40e91323c resourceVersion: "21962" uid: e6a7b00d-b401-493b-9f9f-f1b5bc261941 spec: progressDeadlineSeconds: 600 replicas: 1 revisionHistoryLimit: 10 selector: matchLabels: gateway.networking.k8s.io/gateway-name: router-gateway-1 strategy: rollingUpdate: maxSurge: 25% maxUnavailable: 25% type: RollingUpdate template: metadata: annotations: istio.io/rev: openshift-gateway prometheus.io/path: /stats/prometheus prometheus.io/port: "15020" prometheus.io/scrape: "true" labels: gateway.istio.io/managed: istio.io-gateway-controller gateway.networking.k8s.io/gateway-name: router-gateway-1 service.istio.io/canonical-name: router-gateway-1-openshift-default service.istio.io/canonical-revision: latest sidecar.istio.io/inject: "false" spec: containers: - args: - proxy - router - --domain - $(POD_NAMESPACE).svc.cluster.local - --proxyLogLevel - warning - --proxyComponentLogLevel - misc:error - --log_output_level - default:info env: - name: PILOT_CERT_PROVIDER value: istiod - name: CA_ADDR value: istiod-openshift-gateway.openshift-ingress.svc:15012 - name: POD_NAME valueFrom: fieldRef: apiVersion: v1 fieldPath: metadata.name - name: POD_NAMESPACE valueFrom: fieldRef: apiVersion: v1 fieldPath: metadata.namespace - name: INSTANCE_IP valueFrom: fieldRef: apiVersion: v1 fieldPath: status.podIP - name: SERVICE_ACCOUNT valueFrom: fieldRef: apiVersion: v1 fieldPath: spec.serviceAccountName - name: HOST_IP valueFrom: fieldRef: apiVersion: v1 fieldPath: status.hostIP - name: ISTIO_CPU_LIMIT valueFrom: resourceFieldRef: divisor: "0" resource: limits.cpu - name: PROXY_CONFIG value: | {"discoveryAddress":"istiod-openshift-gateway.openshift-ingress.svc:15012","proxyHeaders":{"server":{"disabled":true},"envoyDebugHeaders":{"disabled":true},"metadataExchangeHeaders":{"mode":"IN_MESH"}}} - name: ISTIO_META_POD_PORTS value: '[]' - name: ISTIO_META_APP_CONTAINERS - name: GOMEMLIMIT valueFrom: resourceFieldRef: divisor: "0" resource: limits.memory - name: GOMAXPROCS valueFrom: resourceFieldRef: divisor: "0" resource: limits.cpu - name: ISTIO_META_CLUSTER_ID value: Kubernetes - name: ISTIO_META_NODE_NAME valueFrom: fieldRef: apiVersion: v1 fieldPath: spec.nodeName - name: ISTIO_META_INTERCEPTION_MODE value: REDIRECT - name: ISTIO_META_WORKLOAD_NAME value: router-gateway-1-openshift-default - name: ISTIO_META_OWNER value: kubernetes://apis/apps/v1/namespaces/kserve-ci-e2e-test/deployments/router-gateway-1-openshift-default - name: ISTIO_META_MESH_ID value: cluster.local - name: TRUST_DOMAIN value: cluster.local image: registry.redhat.io/openshift-service-mesh/istio-proxyv2-rhel9@sha256:7d15cebf9b62f3f235c0eab5158ac8ff2fda86a1d193490dc94c301402c99da8 imagePullPolicy: IfNotPresent name: istio-proxy ports: - containerPort: 15020 name: metrics protocol: TCP - containerPort: 15021 name: status-port protocol: TCP - containerPort: 15090 name: http-envoy-prom protocol: TCP readinessProbe: failureThreshold: 4 httpGet: path: /healthz/ready port: 15021 scheme: HTTP periodSeconds: 15 successThreshold: 1 timeoutSeconds: 1 resources: limits: cpu: "2" memory: 1Gi requests: cpu: 100m memory: 128Mi securityContext: allowPrivilegeEscalation: false capabilities: drop: - ALL privileged: false readOnlyRootFilesystem: true runAsGroup: 1000699999 runAsNonRoot: true runAsUser: 1000699999 startupProbe: failureThreshold: 30 httpGet: path: /healthz/ready port: 15021 scheme: HTTP initialDelaySeconds: 1 periodSeconds: 1 successThreshold: 1 timeoutSeconds: 1 terminationMessagePath: /dev/termination-log terminationMessagePolicy: File volumeMounts: - mountPath: /var/run/secrets/workload-spiffe-uds name: workload-socket - mountPath: /var/run/secrets/credential-uds name: credential-socket - mountPath: /var/run/secrets/workload-spiffe-credentials name: workload-certs - mountPath: /var/run/secrets/istio name: istiod-ca-cert - mountPath: /var/lib/istio/data name: istio-data - mountPath: /etc/istio/proxy name: istio-envoy - mountPath: /var/run/secrets/tokens name: istio-token - mountPath: /etc/istio/pod name: istio-podinfo dnsPolicy: ClusterFirst restartPolicy: Always schedulerName: default-scheduler securityContext: sysctls: - name: net.ipv4.ip_unprivileged_port_start value: "0" serviceAccount: router-gateway-1-openshift-default serviceAccountName: router-gateway-1-openshift-default terminationGracePeriodSeconds: 30 volumes: - emptyDir: {} name: workload-socket - emptyDir: {} name: credential-socket - emptyDir: {} name: workload-certs - emptyDir: medium: Memory name: istio-envoy - emptyDir: {} name: istio-data - downwardAPI: defaultMode: 420 items: - fieldRef: apiVersion: v1 fieldPath: metadata.labels path: labels - fieldRef: apiVersion: v1 fieldPath: metadata.annotations path: annotations name: istio-podinfo - name: istio-token projected: defaultMode: 420 sources: - serviceAccountToken: audience: istio-ca expirationSeconds: 43200 path: istio-token - configMap: defaultMode: 420 name: openshift-gw-ca-root-cert name: istiod-ca-cert status: availableReplicas: 1 conditions: - lastTransitionTime: "2026-07-02T21:31:26Z" lastUpdateTime: "2026-07-02T21:31:29Z" message: ReplicaSet "router-gateway-1-openshift-default-799f46c59b" has successfully progressed. reason: NewReplicaSetAvailable status: "True" type: Progressing - lastTransitionTime: "2026-07-02T21:31:51Z" lastUpdateTime: "2026-07-02T21:31:51Z" message: Deployment has minimum availability. reason: MinimumReplicasAvailable status: "True" type: Available observedGeneration: 1 readyReplicas: 1 replicas: 1 updatedReplicas: 1 - apiVersion: apps/v1 kind: Deployment metadata: annotations: deployment.kubernetes.io/revision: "1" creationTimestamp: "2026-07-02T22:34:48Z" generation: 1 labels: gateway.istio.io/managed: openshift.io-gateway-controller-v1 gateway.networking.k8s.io/gateway-name: router-gateway-2 managedFields: - apiVersion: apps/v1 fieldsType: FieldsV1 fieldsV1: f:metadata: f:annotations: {} f:labels: f:gateway.istio.io/managed: {} f:gateway.networking.k8s.io/gateway-name: {} f:ownerReferences: k:{"uid":"0386c6c7-53c2-4559-80e8-cbc2b907ba18"}: {} f:spec: f:selector: {} f:template: f:metadata: f:annotations: f:istio.io/rev: {} f:prometheus.io/path: {} f:prometheus.io/port: {} f:prometheus.io/scrape: {} f:labels: f:gateway.istio.io/managed: {} f:gateway.networking.k8s.io/gateway-name: {} f:service.istio.io/canonical-name: {} f:service.istio.io/canonical-revision: {} f:sidecar.istio.io/inject: {} f:spec: f:containers: k:{"name":"istio-proxy"}: .: {} f:args: {} f:env: k:{"name":"CA_ADDR"}: .: {} f:name: {} f:value: {} k:{"name":"GOMAXPROCS"}: .: {} f:name: {} f:valueFrom: f:resourceFieldRef: {} k:{"name":"GOMEMLIMIT"}: .: {} f:name: {} f:valueFrom: f:resourceFieldRef: {} k:{"name":"HOST_IP"}: .: {} f:name: {} f:valueFrom: f:fieldRef: {} k:{"name":"INSTANCE_IP"}: .: {} f:name: {} f:valueFrom: f:fieldRef: {} k:{"name":"ISTIO_CPU_LIMIT"}: .: {} f:name: {} f:valueFrom: f:resourceFieldRef: {} k:{"name":"ISTIO_META_APP_CONTAINERS"}: .: {} f:name: {} f:value: {} k:{"name":"ISTIO_META_CLUSTER_ID"}: .: {} f:name: {} f:value: {} k:{"name":"ISTIO_META_INTERCEPTION_MODE"}: .: {} f:name: {} f:value: {} k:{"name":"ISTIO_META_MESH_ID"}: .: {} f:name: {} f:value: {} k:{"name":"ISTIO_META_NODE_NAME"}: .: {} f:name: {} f:valueFrom: f:fieldRef: {} k:{"name":"ISTIO_META_OWNER"}: .: {} f:name: {} f:value: {} k:{"name":"ISTIO_META_POD_PORTS"}: .: {} f:name: {} f:value: {} k:{"name":"ISTIO_META_WORKLOAD_NAME"}: .: {} f:name: {} f:value: {} k:{"name":"PILOT_CERT_PROVIDER"}: .: {} f:name: {} f:value: {} k:{"name":"POD_NAME"}: .: {} f:name: {} f:valueFrom: f:fieldRef: {} k:{"name":"POD_NAMESPACE"}: .: {} f:name: {} f:valueFrom: f:fieldRef: {} k:{"name":"PROXY_CONFIG"}: .: {} f:name: {} f:value: {} k:{"name":"SERVICE_ACCOUNT"}: .: {} f:name: {} f:valueFrom: f:fieldRef: {} k:{"name":"TRUST_DOMAIN"}: .: {} f:name: {} f:value: {} f:image: {} f:name: {} f:ports: k:{"containerPort":15020,"protocol":"TCP"}: .: {} f:containerPort: {} f:name: {} f:protocol: {} k:{"containerPort":15021,"protocol":"TCP"}: .: {} f:containerPort: {} f:name: {} f:protocol: {} k:{"containerPort":15090,"protocol":"TCP"}: .: {} f:containerPort: {} f:name: {} f:protocol: {} f:readinessProbe: f:failureThreshold: {} f:httpGet: f:path: {} f:port: {} f:scheme: {} 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:privileged: {} f:readOnlyRootFilesystem: {} f:runAsGroup: {} f:runAsNonRoot: {} f:runAsUser: {} f:startupProbe: f:failureThreshold: {} f:httpGet: f:path: {} f:port: {} f:scheme: {} f:initialDelaySeconds: {} f:periodSeconds: {} f:successThreshold: {} f:timeoutSeconds: {} f:volumeMounts: k:{"mountPath":"/etc/istio/pod"}: .: {} f:mountPath: {} f:name: {} k:{"mountPath":"/etc/istio/proxy"}: .: {} f:mountPath: {} f:name: {} k:{"mountPath":"/var/lib/istio/data"}: .: {} f:mountPath: {} f:name: {} k:{"mountPath":"/var/run/secrets/credential-uds"}: .: {} f:mountPath: {} f:name: {} k:{"mountPath":"/var/run/secrets/istio"}: .: {} f:mountPath: {} f:name: {} k:{"mountPath":"/var/run/secrets/tokens"}: .: {} f:mountPath: {} f:name: {} k:{"mountPath":"/var/run/secrets/workload-spiffe-credentials"}: .: {} f:mountPath: {} f:name: {} k:{"mountPath":"/var/run/secrets/workload-spiffe-uds"}: .: {} f:mountPath: {} f:name: {} f:securityContext: f:sysctls: {} f:serviceAccountName: {} f:volumes: k:{"name":"credential-socket"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"istio-data"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"istio-envoy"}: .: {} f:emptyDir: f:medium: {} f:name: {} k:{"name":"istio-podinfo"}: .: {} f:downwardAPI: f:items: {} f:name: {} k:{"name":"istio-token"}: .: {} f:name: {} f:projected: f:sources: {} k:{"name":"istiod-ca-cert"}: .: {} f:configMap: f:name: {} f:name: {} k:{"name":"workload-certs"}: .: {} f:emptyDir: {} f:name: {} k:{"name":"workload-socket"}: .: {} f:emptyDir: {} f:name: {} manager: openshift.io/gateway-controller/v1 operation: Apply time: "2026-07-02T22:34:48Z" - 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:updatedReplicas: {} manager: kube-controller-manager operation: Update subresource: status time: "2026-07-02T22:35:10Z" name: router-gateway-2-openshift-default namespace: kserve-ci-e2e-test ownerReferences: - apiVersion: gateway.networking.k8s.io/v1beta1 kind: Gateway name: router-gateway-2 uid: 0386c6c7-53c2-4559-80e8-cbc2b907ba18 resourceVersion: "80278" uid: 5aa705a5-073d-4544-bef2-d93fe5a915e0 spec: progressDeadlineSeconds: 600 replicas: 1 revisionHistoryLimit: 10 selector: matchLabels: gateway.networking.k8s.io/gateway-name: router-gateway-2 strategy: rollingUpdate: maxSurge: 25% maxUnavailable: 25% type: RollingUpdate template: metadata: annotations: istio.io/rev: openshift-gateway prometheus.io/path: /stats/prometheus prometheus.io/port: "15020" prometheus.io/scrape: "true" labels: gateway.istio.io/managed: istio.io-gateway-controller gateway.networking.k8s.io/gateway-name: router-gateway-2 service.istio.io/canonical-name: router-gateway-2-openshift-default service.istio.io/canonical-revision: latest sidecar.istio.io/inject: "false" spec: containers: - args: - proxy - router - --domain - $(POD_NAMESPACE).svc.cluster.local - --proxyLogLevel - warning - --proxyComponentLogLevel - misc:error - --log_output_level - default:info env: - name: PILOT_CERT_PROVIDER value: istiod - name: CA_ADDR value: istiod-openshift-gateway.openshift-ingress.svc:15012 - name: POD_NAME valueFrom: fieldRef: apiVersion: v1 fieldPath: metadata.name - name: POD_NAMESPACE valueFrom: fieldRef: apiVersion: v1 fieldPath: metadata.namespace - name: INSTANCE_IP valueFrom: fieldRef: apiVersion: v1 fieldPath: status.podIP - name: SERVICE_ACCOUNT valueFrom: fieldRef: apiVersion: v1 fieldPath: spec.serviceAccountName - name: HOST_IP valueFrom: fieldRef: apiVersion: v1 fieldPath: status.hostIP - name: ISTIO_CPU_LIMIT valueFrom: resourceFieldRef: divisor: "0" resource: limits.cpu - name: PROXY_CONFIG value: | {"discoveryAddress":"istiod-openshift-gateway.openshift-ingress.svc:15012","proxyHeaders":{"server":{"disabled":true},"envoyDebugHeaders":{"disabled":true},"metadataExchangeHeaders":{"mode":"IN_MESH"}}} - name: ISTIO_META_POD_PORTS value: '[]' - name: ISTIO_META_APP_CONTAINERS - name: GOMEMLIMIT valueFrom: resourceFieldRef: divisor: "0" resource: limits.memory - name: GOMAXPROCS valueFrom: resourceFieldRef: divisor: "0" resource: limits.cpu - name: ISTIO_META_CLUSTER_ID value: Kubernetes - name: ISTIO_META_NODE_NAME valueFrom: fieldRef: apiVersion: v1 fieldPath: spec.nodeName - name: ISTIO_META_INTERCEPTION_MODE value: REDIRECT - name: ISTIO_META_WORKLOAD_NAME value: router-gateway-2-openshift-default - name: ISTIO_META_OWNER value: kubernetes://apis/apps/v1/namespaces/kserve-ci-e2e-test/deployments/router-gateway-2-openshift-default - name: ISTIO_META_MESH_ID value: cluster.local - name: TRUST_DOMAIN value: cluster.local image: registry.redhat.io/openshift-service-mesh/istio-proxyv2-rhel9@sha256:7d15cebf9b62f3f235c0eab5158ac8ff2fda86a1d193490dc94c301402c99da8 imagePullPolicy: IfNotPresent name: istio-proxy ports: - containerPort: 15020 name: metrics protocol: TCP - containerPort: 15021 name: status-port protocol: TCP - containerPort: 15090 name: http-envoy-prom protocol: TCP readinessProbe: failureThreshold: 4 httpGet: path: /healthz/ready port: 15021 scheme: HTTP periodSeconds: 15 successThreshold: 1 timeoutSeconds: 1 resources: limits: cpu: "2" memory: 1Gi requests: cpu: 100m memory: 128Mi securityContext: allowPrivilegeEscalation: false capabilities: drop: - ALL privileged: false readOnlyRootFilesystem: true runAsGroup: 1000699999 runAsNonRoot: true runAsUser: 1000699999 startupProbe: failureThreshold: 30 httpGet: path: /healthz/ready port: 15021 scheme: HTTP initialDelaySeconds: 1 periodSeconds: 1 successThreshold: 1 timeoutSeconds: 1 terminationMessagePath: /dev/termination-log terminationMessagePolicy: File volumeMounts: - mountPath: /var/run/secrets/workload-spiffe-uds name: workload-socket - mountPath: /var/run/secrets/credential-uds name: credential-socket - mountPath: /var/run/secrets/workload-spiffe-credentials name: workload-certs - mountPath: /var/run/secrets/istio name: istiod-ca-cert - mountPath: /var/lib/istio/data name: istio-data - mountPath: /etc/istio/proxy name: istio-envoy - mountPath: /var/run/secrets/tokens name: istio-token - mountPath: /etc/istio/pod name: istio-podinfo dnsPolicy: ClusterFirst restartPolicy: Always schedulerName: default-scheduler securityContext: sysctls: - name: net.ipv4.ip_unprivileged_port_start value: "0" serviceAccount: router-gateway-2-openshift-default serviceAccountName: router-gateway-2-openshift-default terminationGracePeriodSeconds: 30 volumes: - emptyDir: {} name: workload-socket - emptyDir: {} name: credential-socket - emptyDir: {} name: workload-certs - emptyDir: medium: Memory name: istio-envoy - emptyDir: {} name: istio-data - downwardAPI: defaultMode: 420 items: - fieldRef: apiVersion: v1 fieldPath: metadata.labels path: labels - fieldRef: apiVersion: v1 fieldPath: metadata.annotations path: annotations name: istio-podinfo - name: istio-token projected: defaultMode: 420 sources: - serviceAccountToken: audience: istio-ca expirationSeconds: 43200 path: istio-token - configMap: defaultMode: 420 name: openshift-gw-ca-root-cert name: istiod-ca-cert status: availableReplicas: 1 conditions: - lastTransitionTime: "2026-07-02T22:34:48Z" lastUpdateTime: "2026-07-02T22:34:53Z" message: ReplicaSet "router-gateway-2-openshift-default-54c789bdc6" has successfully progressed. reason: NewReplicaSetAvailable status: "True" type: Progressing - lastTransitionTime: "2026-07-02T22:35:10Z" lastUpdateTime: "2026-07-02T22:35:10Z" message: Deployment has minimum availability. reason: MinimumReplicasAvailable status: "True" type: Available observedGeneration: 1 readyReplicas: 1 replicas: 1 updatedReplicas: 1 kind: DeploymentList metadata: resourceVersion: "108289"