Dra support

helm/README.md

@@ -9,20 +9,8 @@ This documentation contains documentation for the NV-Ingest Helm charts.
 
 Before you install the Helm charts, be sure you meet the hardware and software prerequisites. Refer to the [supported configurations](https://github.com/NVIDIA/nv-ingest?tab=readme-ov-file#hardware).
 
-> Starting with version 26.1.0, the [NVIDIA NIM Operator](https://docs.nvidia.com/nim-operator/latest/install.html) is required. All NIM services are now deployed by using NIM Operator CRDs (NIMCache and NIMService), not Helm subcharts.
->
-> **Upgrading from 25.9.0:**
-> 1. Install NIM Operator before upgrading
-> 2. Update your values file with the new configuration keys:
->
-> | 25.9.0 | 26.x |
-> |--------|------|
-> | `nim-vlm-image-captioning.deployed=true` | `nimOperator.nemotron_nano_12b_v2_vl.enabled=true` |
-> | `paddleocr-nim.deployed=true` | `nimOperator.nemoretriever_ocr_v1.enabled=true` |
-> | `riva-nim.deployed=true` | `nimOperator.audio.enabled=true` |
-> | `nim-vlm-text-extraction.deployed=true` | `nimOperator.nemotron_parse.enabled=true` |
-
-The [Nvidia GPU Operator](https://docs.nvidia.com/datacenter/cloud-native/gpu-operator/latest/getting-started.html) must also be installed and configured in your cluster.
+The [Nvidia nim-operator](https://docs.nvidia.com/nim-operator/latest/install.html) must also be installed and configured in your cluster to ensure that
+the Nvidia NIMs are properly deployed.
 
 ## Initial Environment Setup
 
@@ -102,6 +90,52 @@ In this case, make sure to remove the following from your helm command:
     --set ngcApiSecret.password="${NGC_API_KEY}" \
 ```
 
+### Install or Upgrade the Helm Chart with NVIDIA DRA Support
+
+- Prerequisites: 1 GPU with at least 64GB memory
+
+create a `custom-values.yaml` with below content
+
+```yaml
+nv-ingest:
+  nimOperator:
+    draResources:
+      enabled: true
+      name: "nvingest-claim"
+    embedqa:
+      draResources:
+      - resourceClaimName: nvingest-claim
+    page_elements:
+      draResources:
+      - resourceClaimName: nvingest-claim
+    graphic_elements:
+      draResources:
+      - resourceClaimName: nvingest-claim
+    table_structure:
+      draResources:
+      - resourceClaimName: nvingest-claim
+    nemoretriever_ocr_v1:
+      draResources:
+      - resourceClaimName: nvingest-claim
+```
+To install or upgrade the Helm chart with [NVIDIA DRA](https://docs.nvidia.com/datacenter/cloud-native/gpu-operator/latest/dra-intro-install.html), run the following code.
+```bash
+helm upgrade \
+    --install \
+    nv-ingest \
+    https://helm.ngc.nvidia.com/nvidia/nemo-microservices/charts/nv-ingest-25.9.0.tgz \
+    -n ${NAMESPACE} \
+    --username '$oauthtoken' \
+    --password "${NGC_API_KEY}" \
+    --set ngcImagePullSecret.create=true \
+    --set ngcImagePullSecret.password="${NGC_API_KEY}" \
+    --set ngcApiSecret.create=true \
+    --set ngcApiSecret.password="${NGC_API_KEY}" \
+    --set image.repository="nvcr.io/nvidia/nemo-microservices/nv-ingest" \
+    --set image.tag="25.9.0" \
+    -f ./custom-values.yaml
+```
+
 ## Usage
 
 Jobs are submitted via the `nv-ingest-cli` command.
@@ -428,7 +462,9 @@ You can also use NV-Ingest's Python client API to interact with the service runn
 | nimOperator.audio.resources.limits."nvidia.com/gpu" | int | `1` |  |
 | nimOperator.audio.storage.pvc.create | bool | `true` |  |
 | nimOperator.audio.storage.pvc.size | string | `"25Gi"` |  |
-| nimOperator.audio.storage.pvc.volumeAccessMode | string | `"ReadWriteMany"` |  |
+| nimOperator.audio.storage.pvc.volumeAccessMode | string | `"ReadWriteOnce"` |  |
+| nimOperator.draResources.enabled | bool | `false` |  |
+| nimOperator.draResources.name | string | `"nvingest-claim"` |  |
 | nimOperator.embedqa.authSecret | string | `"ngc-api"` |  |
 | nimOperator.embedqa.enabled | bool | `true` |  |
 | nimOperator.embedqa.env[0].name | string | `"NIM_HTTP_API_PORT"` |  |
@@ -445,12 +481,12 @@ You can also use NV-Ingest's Python client API to interact with the service runn
 | nimOperator.embedqa.image.pullPolicy | string | `"IfNotPresent"` |  |
 | nimOperator.embedqa.image.pullSecrets[0] | string | `"ngc-secret"` |  |
 | nimOperator.embedqa.image.repository | string | `"nvcr.io/nim/nvidia/llama-3.2-nv-embedqa-1b-v2"` |  |
-| nimOperator.embedqa.image.tag | string | `"1.10.0"` |  |
+| nimOperator.embedqa.image.tag | string | `"1.10.1"` |  |
 | nimOperator.embedqa.replicas | int | `1` |  |
 | nimOperator.embedqa.resources.limits."nvidia.com/gpu" | int | `1` |  |
 | nimOperator.embedqa.storage.pvc.create | bool | `true` |  |
 | nimOperator.embedqa.storage.pvc.size | string | `"50Gi"` |  |
-| nimOperator.embedqa.storage.pvc.volumeAccessMode | string | `"ReadWriteMany"` |  |
+| nimOperator.embedqa.storage.pvc.volumeAccessMode | string | `"ReadWriteOnce"` |  |
 | nimOperator.graphic_elements.authSecret | string | `"ngc-api"` |  |
 | nimOperator.graphic_elements.enabled | bool | `true` |  |
 | nimOperator.graphic_elements.env[0].name | string | `"NIM_HTTP_API_PORT"` |  |
@@ -476,7 +512,7 @@ You can also use NV-Ingest's Python client API to interact with the service runn
 | nimOperator.graphic_elements.resources.limits."nvidia.com/gpu" | int | `1` |  |
 | nimOperator.graphic_elements.storage.pvc.create | bool | `true` |  |
 | nimOperator.graphic_elements.storage.pvc.size | string | `"25Gi"` |  |
-| nimOperator.graphic_elements.storage.pvc.volumeAccessMode | string | `"ReadWriteMany"` |  |
+| nimOperator.graphic_elements.storage.pvc.volumeAccessMode | string | `"ReadWriteOnce"` |  |
 | nimOperator.llama_3_2_nv_rerankqa_1b_v2.authSecret | string | `"ngc-api"` |  |
 | nimOperator.llama_3_2_nv_rerankqa_1b_v2.enabled | bool | `false` |  |
 | nimOperator.llama_3_2_nv_rerankqa_1b_v2.env | list | `[]` |  |
@@ -491,7 +527,7 @@ You can also use NV-Ingest's Python client API to interact with the service runn
 | nimOperator.llama_3_2_nv_rerankqa_1b_v2.resources.limits."nvidia.com/gpu" | int | `1` |  |
 | nimOperator.llama_3_2_nv_rerankqa_1b_v2.storage.pvc.create | bool | `true` |  |
 | nimOperator.llama_3_2_nv_rerankqa_1b_v2.storage.pvc.size | string | `"50Gi"` |  |
-| nimOperator.llama_3_2_nv_rerankqa_1b_v2.storage.pvc.volumeAccessMode | string | `"ReadWriteMany"` |  |
+| nimOperator.llama_3_2_nv_rerankqa_1b_v2.storage.pvc.volumeAccessMode | string | `"ReadWriteOnce"` |  |
 | nimOperator.nemoretriever_ocr_v1.authSecret | string | `"ngc-api"` |  |
 | nimOperator.nemoretriever_ocr_v1.enabled | bool | `true` |  |
 | nimOperator.nemoretriever_ocr_v1.env[0].name | string | `"OMP_NUM_THREADS"` |  |
@@ -519,7 +555,7 @@ You can also use NV-Ingest's Python client API to interact with the service runn
 | nimOperator.nemoretriever_ocr_v1.resources.limits."nvidia.com/gpu" | int | `1` |  |
 | nimOperator.nemoretriever_ocr_v1.storage.pvc.create | bool | `true` |  |
 | nimOperator.nemoretriever_ocr_v1.storage.pvc.size | string | `"25Gi"` |  |
-| nimOperator.nemoretriever_ocr_v1.storage.pvc.volumeAccessMode | string | `"ReadWriteMany"` |  |
+| nimOperator.nemoretriever_ocr_v1.storage.pvc.volumeAccessMode | string | `"ReadWriteOnce"` |  |
 | nimOperator.nemotron_nano_12b_v2_vl.authSecret | string | `"ngc-api"` |  |
 | nimOperator.nemotron_nano_12b_v2_vl.enabled | bool | `false` |  |
 | nimOperator.nemotron_nano_12b_v2_vl.expose.service.grpcPort | int | `8001` |  |
@@ -533,7 +569,7 @@ You can also use NV-Ingest's Python client API to interact with the service runn
 | nimOperator.nemotron_nano_12b_v2_vl.resources.limits."nvidia.com/gpu" | int | `1` |  |
 | nimOperator.nemotron_nano_12b_v2_vl.storage.pvc.create | bool | `true` |  |
 | nimOperator.nemotron_nano_12b_v2_vl.storage.pvc.size | string | `"300Gi"` |  |
-| nimOperator.nemotron_nano_12b_v2_vl.storage.pvc.volumeAccessMode | string | `"ReadWriteMany"` |  |
+| nimOperator.nemotron_nano_12b_v2_vl.storage.pvc.volumeAccessMode | string | `"ReadWriteOnce"` |  |
 | nimOperator.nemotron_parse.authSecret | string | `"ngc-api"` |  |
 | nimOperator.nemotron_parse.enabled | bool | `false` |  |
 | nimOperator.nemotron_parse.env[0].name | string | `"NIM_HTTP_API_PORT"` |  |
@@ -551,11 +587,11 @@ You can also use NV-Ingest's Python client API to interact with the service runn
 | nimOperator.nemotron_parse.resources.limits."nvidia.com/gpu" | int | `1` |  |
 | nimOperator.nemotron_parse.storage.pvc.create | bool | `true` |  |
 | nimOperator.nemotron_parse.storage.pvc.size | string | `"100Gi"` |  |
-| nimOperator.nemotron_parse.storage.pvc.volumeAccessMode | string | `"ReadWriteMany"` |  |
+| nimOperator.nemotron_parse.storage.pvc.volumeAccessMode | string | `"ReadWriteOnce"` |  |
 | nimOperator.nimCache.pvc.create | bool | `true` |  |
 | nimOperator.nimCache.pvc.size | string | `"25Gi"` |  |
 | nimOperator.nimCache.pvc.storageClass | string | `"default"` |  |
-| nimOperator.nimCache.pvc.volumeAccessMode | string | `"ReadWriteMany"` |  |
+| nimOperator.nimCache.pvc.volumeAccessMode | string | `"ReadWriteOnce"` |  |
 | nimOperator.nimService.namespaces | list | `[]` |  |
 | nimOperator.nimService.resources | object | `{}` |  |
 | nimOperator.page_elements.authSecret | string | `"ngc-api"` |  |
@@ -603,7 +639,7 @@ You can also use NV-Ingest's Python client API to interact with the service runn
 | nimOperator.page_elements.resources.limits."nvidia.com/gpu" | int | `1` |  |
 | nimOperator.page_elements.storage.pvc.create | bool | `true` |  |
 | nimOperator.page_elements.storage.pvc.size | string | `"25Gi"` |  |
-| nimOperator.page_elements.storage.pvc.volumeAccessMode | string | `"ReadWriteMany"` |  |
+| nimOperator.page_elements.storage.pvc.volumeAccessMode | string | `"ReadWriteOnce"` |  |
 | nimOperator.table_structure.authSecret | string | `"ngc-api"` |  |
 | nimOperator.table_structure.enabled | bool | `true` |  |
 | nimOperator.table_structure.env[0].name | string | `"NIM_HTTP_API_PORT"` |  |
@@ -629,7 +665,7 @@ You can also use NV-Ingest's Python client API to interact with the service runn
 | nimOperator.table_structure.resources.limits."nvidia.com/gpu" | int | `1` |  |
 | nimOperator.table_structure.storage.pvc.create | bool | `true` |  |
 | nimOperator.table_structure.storage.pvc.size | string | `"25Gi"` |  |
-| nimOperator.table_structure.storage.pvc.volumeAccessMode | string | `"ReadWriteMany"` |  |
+| nimOperator.table_structure.storage.pvc.volumeAccessMode | string | `"ReadWriteOnce"` |  |
 | nodeSelector | object | `{}` |  |
 | opentelemetry-collector.config.exporters.debug.verbosity | string | `"detailed"` |  |
 | opentelemetry-collector.config.exporters.zipkin.endpoint | string | `"http://nv-ingest-zipkin:9411/api/v2/spans"` |  |
@@ -682,7 +718,7 @@ You can also use NV-Ingest's Python client API to interact with the service runn
 | prometheus.alertmanager.enabled | bool | `false` |  |
 | prometheus.enabled | bool | `false` |  |
 | prometheus.server.enabled | bool | `false` |  |
-| readinessProbe.enabled | bool | `false` |  |
+| readinessProbe.enabled | bool | `true` |  |
 | readinessProbe.failureThreshold | int | `220` |  |
 | readinessProbe.httpGet.path | string | `"/v1/health/ready"` |  |
 | readinessProbe.httpGet.port | string | `"http"` |  |

helm/README.md.gotmpl

@@ -92,6 +92,52 @@ In this case, make sure to remove the following from your helm command:
     --set ngcApiSecret.password="${NGC_API_KEY}" \
 ```
 
+### Install or Upgrade the Helm Chart with NVIDIA DRA Support
+
+- Prerequisites: 1 GPU with at least 64GB memory
+
+create a `custom-values.yaml` with below content
+
+```yaml
+nv-ingest:
+  nimOperator:
+    draResources:
+      enabled: true
+      name: "nvingest-claim"
+    embedqa:
+      draResources:
+      - resourceClaimName: nvingest-claim
+    page_elements:
+      draResources:
+      - resourceClaimName: nvingest-claim
+    graphic_elements:
+      draResources:
+      - resourceClaimName: nvingest-claim
+    table_structure:
+      draResources:
+      - resourceClaimName: nvingest-claim
+    nemoretriever_ocr_v1:
+      draResources:
+      - resourceClaimName: nvingest-claim
+```
+To install or upgrade the Helm chart with [NVIDIA DRA](https://docs.nvidia.com/datacenter/cloud-native/gpu-operator/latest/dra-intro-install.html), run the following code.
+```bash
+helm upgrade \
+    --install \
+    nv-ingest \
+    https://helm.ngc.nvidia.com/nvidia/nemo-microservices/charts/nv-ingest-25.9.0.tgz \
+    -n ${NAMESPACE} \
+    --username '$oauthtoken' \
+    --password "${NGC_API_KEY}" \
+    --set ngcImagePullSecret.create=true \
+    --set ngcImagePullSecret.password="${NGC_API_KEY}" \
+    --set ngcApiSecret.create=true \
+    --set ngcApiSecret.password="${NGC_API_KEY}" \
+    --set image.repository="nvcr.io/nvidia/nemo-microservices/nv-ingest" \
+    --set image.tag="25.9.0" \
+    -f ./custom-values.yaml
+```
+
 ## Usage
 
 Jobs are submitted via the `nv-ingest-cli` command.

helm/templates/deployment.yaml

@@ -145,8 +145,15 @@ spec:
           readinessProbe:
             {{- toYaml (omit .Values.readinessProbe "enabled") | nindent 12 }}
           {{- end }}
+          {{- if .Values.nimOperator.draResources.enabled }}
           resources:
-            {{- toYaml .Values.resources | nindent 12 }}
+            claims:
+            - name: shared-gpu
+          {{- end }}
+          {{- if not .Values.nimOperator.draResources.enabled }}
+          resources:
+            {{ toYaml .Values.resources | nindent 12 }}
+          {{- end }}
           volumeMounts:
             - mountPath: /dev/shm
               name: dshm
@@ -163,6 +170,11 @@ spec:
       nodeSelector:
         {{- toYaml . | nindent 8 }}
       {{- end }}
+      {{- if .Values.nimOperator.draResources.enabled }}
+      resourceClaims:
+      - name: shared-gpu
+        resourceClaimName: {{ .Values.nimOperator.draResources.name }}
+      {{- end }}
       {{- with .Values.affinity }}
       affinity:
         {{- toYaml . | nindent 8 }}
@@ -183,4 +195,4 @@ spec:
         - name: {{ $k }}
           {{- toYaml $v | nindent 10 }}
       {{- end }}
-      {{- end }}
+      {{- end }}

helm/templates/dra.yaml

@@ -0,0 +1,23 @@
+{{- $claimName := .Values.nimOperator.draResources.name -}}
+{{- $existing := lookup "resource.k8s.io/v1beta2" "ResourceClaim" .Release.Namespace $claimName -}}
+{{- if and
+      (.Capabilities.APIVersions.Has "resource.k8s.io/v1beta2")
+      (.Values.nimOperator.draResources.enabled)
+      (not $existing)
+}}
+apiVersion: resource.k8s.io/v1beta2
+kind: ResourceClaim
+metadata:
+  name: {{ $claimName }}
+  annotations:
+    "helm.sh/hook": pre-install
+    "helm.sh/hook-weight": "5"
+spec:
+  devices:
+    requests:
+      - name: {{ $claimName }}-gpu
+        exactly:
+          deviceClassName: gpu.nvidia.com
+          count: 1
+          allocationMode: ExactCount
+{{- end }}

helm/templates/llama-3.2-nv-embedqa-1b-v2.yaml

@@ -36,12 +36,18 @@ spec:
   replicas: {{ .Values.nimOperator.embedqa.replicas }}
   nodeSelector:
 {{ toYaml .Values.nimOperator.embedqa.nodeSelector | nindent 4 }}
+{{- if .Values.nimOperator.draResources.enabled }}
+  draResources:
+{{ toYaml .Values.nimOperator.llama_3_2_nv_rerankqa_1b_v2.draResources | nindent 4 }}
+{{- end }}
+{{- if not .Values.nimOperator.draResources.enabled }}
   resources:
-{{ toYaml .Values.nimOperator.embedqa.resources | nindent 4 }}
+{{ toYaml .Values.nimOperator.llama_3_2_nv_rerankqa_1b_v2.resources | nindent 4 }}
+{{- end }}
   tolerations:
 {{ toYaml .Values.nimOperator.embedqa.tolerations | nindent 4 }}
   expose:
 {{ toYaml .Values.nimOperator.embedqa.expose | nindent 4 }}
   env:
 {{ toYaml .Values.nimOperator.embedqa.env | nindent 4 }}
-{{- end }}
+{{- end }}

helm/templates/llama-3.2-nv-rerankqa-1b-v2.yaml

@@ -36,8 +36,14 @@ spec:
   replicas: {{ .Values.nimOperator.llama_3_2_nv_rerankqa_1b_v2.replicas }}
   nodeSelector:
 {{ toYaml .Values.nimOperator.llama_3_2_nv_rerankqa_1b_v2.nodeSelector | nindent 4 }}
+{{- if .Values.nimOperator.draResources.enabled }}
+  draResources:
+{{ toYaml .Values.nimOperator.llama_3_2_nv_rerankqa_1b_v2.draResources | nindent 4 }}
+{{- end }}
+{{- if not .Values.nimOperator.draResources.enabled }}
   resources:
 {{ toYaml .Values.nimOperator.llama_3_2_nv_rerankqa_1b_v2.resources | nindent 4 }}
+{{- end }}
   tolerations:
 {{ toYaml .Values.nimOperator.llama_3_2_nv_rerankqa_1b_v2.tolerations | nindent 4 }}
   expose:

helm/templates/nemoretriever-graphic-elements-v1.yaml

@@ -36,8 +36,14 @@ spec:
   replicas: {{ .Values.nimOperator.graphic_elements.replicas }}
   nodeSelector:
 {{ toYaml .Values.nimOperator.graphic_elements.nodeSelector | nindent 4 }}
+{{- if .Values.nimOperator.draResources.enabled }}
+  draResources:
+{{ toYaml .Values.nimOperator.graphic_elements.draResources | nindent 4 }}
+{{- end }}
+{{- if not .Values.nimOperator.draResources.enabled }}
   resources:
 {{ toYaml .Values.nimOperator.graphic_elements.resources | nindent 4 }}
+{{- end }}
   tolerations:
 {{ toYaml .Values.nimOperator.graphic_elements.tolerations | nindent 4 }}
   expose:

helm/templates/nemoretriever-ocr-v1.yaml

@@ -34,7 +34,14 @@ spec:
       name: nemoretriever-ocr-v1
   replicas: {{ .Values.nimOperator.nemoretriever_ocr_v1.replicas }}
   nodeSelector: {{ toYaml .Values.nimOperator.nemoretriever_ocr_v1.nodeSelector | nindent 4 }}
-  resources: {{ toYaml .Values.nimOperator.nemoretriever_ocr_v1.resources | nindent 4 }}
+{{- if .Values.nimOperator.draResources.enabled }}
+  draResources:
+{{ toYaml .Values.nimOperator.nemoretriever_ocr_v1.draResources | nindent 4 }}
+{{- end }}
+{{- if not .Values.nimOperator.draResources.enabled }}
+  resources:
+{{ toYaml .Values.nimOperator.nemoretriever_ocr_v1.resources | nindent 4 }}
+{{- end }}
   tolerations: {{ toYaml .Values.nimOperator.nemoretriever_ocr_v1.tolerations | nindent 4 }}
   expose: {{ toYaml .Values.nimOperator.nemoretriever_ocr_v1.expose | nindent 4 }}
   env: {{ toYaml .Values.nimOperator.nemoretriever_ocr_v1.env | nindent 4 }}

helm/templates/nemoretriever-page-elements-v3.yaml

@@ -36,7 +36,14 @@ spec:
       profile: ''
   replicas: {{ .Values.nimOperator.page_elements.replicas }}
   nodeSelector: {{ toYaml .Values.nimOperator.page_elements.nodeSelector | nindent 4 }}
-  resources: {{ toYaml .Values.nimOperator.page_elements.resources | nindent 4 }}
+{{- if .Values.nimOperator.draResources.enabled }}
+  draResources:
+{{ toYaml .Values.nimOperator.page_elements.draResources | nindent 4 }}
+{{- end }}
+{{- if not .Values.nimOperator.draResources.enabled }}
+  resources:
+{{ toYaml .Values.nimOperator.page_elements.resources | nindent 4 }}
+{{- end }}
   tolerations: {{ toYaml .Values.nimOperator.page_elements.tolerations | nindent 4 }}
   expose: {{ toYaml .Values.nimOperator.page_elements.expose | nindent 4 }}
   env: {{ toYaml .Values.nimOperator.page_elements.env | nindent 4 }}

helm/templates/nemoretriever-table-structure-v1.yaml

@@ -36,8 +36,14 @@ spec:
   replicas: {{ .Values.nimOperator.table_structure.replicas }}
   nodeSelector:
 {{ toYaml .Values.nimOperator.table_structure.nodeSelector | nindent 4 }}
+{{- if .Values.nimOperator.draResources.enabled }}
+  draResources:
+{{ toYaml .Values.nimOperator.table_structure.draResources | nindent 4 }}
+{{- end }}
+{{- if not .Values.nimOperator.draResources.enabled }}
   resources:
 {{ toYaml .Values.nimOperator.table_structure.resources | nindent 4 }}
+{{- end }}
   tolerations:
 {{ toYaml .Values.nimOperator.table_structure.tolerations | nindent 4 }}
   expose: