Merge pull request #1127 from NVIDIA/cpicks-release-24.9-2

[release-24.9] cherrypick changes for release 24.9.1
NVIDIA · Nov 21, 2024 · 57a1009 · 57a1009
2 parents 373b6a8 + 84994be
commit 57a1009
Show file tree

Hide file tree

Showing 6 changed files with 26 additions and 24 deletions.
diff --git a/.github/workflows/ci.yaml b/.github/workflows/ci.yaml
@@ -15,15 +15,9 @@
 name: CI
 
 on:
-  pull_request:
-    types:
-      - opened
-      - synchronize
-    branches:
-      - main
-      - release-*
   push:
     branches:
+      - "pull-request/[0-9]+"
       - main
       - release-*
 
@@ -213,7 +207,7 @@ jobs:
   ### e2e tests ###
   e2e-tests-containerd:
     needs: [build-gpu-operator, build-gpu-operator-validator]
-    runs-on: ubuntu-latest
+    runs-on: linux-amd64-cpu4
     steps:
       - uses: actions/checkout@v4
         name: Check out code
@@ -265,7 +259,7 @@ jobs:
 
   e2e-tests-nvidiadriver:
     needs: [build-gpu-operator, build-gpu-operator-validator]
-    runs-on: ubuntu-latest
+    runs-on: linux-amd64-cpu4
     steps:
       - uses: actions/checkout@v4
         name: Check out code

diff --git a/.nvidia-ci.yml b/.nvidia-ci.yml
@@ -237,7 +237,7 @@ sign:ngc-gpu-operator-validator:
     OPERATOR_IMAGE: "${STAGING_REGISTRY}/gpu-operator"
     VALIDATOR_VERSION: "${CI_COMMIT_SHORT_SHA}"
     VALIDATOR_IMAGE: "${STAGING_REGISTRY}/gpu-operator-validator"
-    TARGET_DRIVER_VERSION: "525.147.05"
+    TARGET_DRIVER_VERSION: "565.57.01"
 
 .e2e_tests:
   extends:

diff --git a/bundle/manifests/gpu-operator-certified.clusterserviceversion.yaml b/bundle/manifests/gpu-operator-certified.clusterserviceversion.yaml
@@ -161,7 +161,7 @@ metadata:
             "driverType": "gpu",
             "repository": "nvcr.io/nvidia",
             "image": "driver",
-            "version": "sha256:8a9a9e9470f64d340a7f3106a03e9622fa98e25368fbfb7ce9c416ad98f6d951",
+            "version": "sha256:78e00fc8aa72f5f4925d54b4f9fbb725ca2168e890d90eadd6a497b7f31ccc18",
             "nodeSelector": {},
             "manager": {},
             "repoConfig": {
@@ -203,15 +203,15 @@ spec:
     - name: gpu-operator-image
       image: registry.gitlab.com/nvidia/kubernetes/gpu-operator/staging/gpu-operator:main-latest
     - name: dcgm-exporter-image
-      image: nvcr.io/nvidia/k8s/dcgm-exporter@sha256:10ff95e83bc137796f5be76278a6b38fd31c7360e62c7e72011b428f8848c791
+      image: nvcr.io/nvidia/k8s/dcgm-exporter@sha256:857f7669ae5ce2a43a3b5691fa970085098321ef8e90a896fa8dda3dee5b1d2b
     - name: dcgm-image
-      image: nvcr.io/nvidia/cloud-native/dcgm@sha256:7b0ebd6c40a11b6484dc4385605372511e4e93132a44d2a3d6ec2e36c24e6783
+      image: nvcr.io/nvidia/cloud-native/dcgm@sha256:e8398b2451996e5c64fc3855e7a3c93b667a85f69c0e259049d3b0012c21545d
     - name: container-toolkit-image
       image: nvcr.io/nvidia/k8s/container-toolkit@sha256:f9d5652cba9ff27f9c4d17f5d09b4a5e4aa631b5cabb9cedf4d3e8c43d6847a4
     - name: driver-image
-      image: nvcr.io/nvidia/driver@sha256:8a9a9e9470f64d340a7f3106a03e9622fa98e25368fbfb7ce9c416ad98f6d951
+      image: nvcr.io/nvidia/driver@sha256:78e00fc8aa72f5f4925d54b4f9fbb725ca2168e890d90eadd6a497b7f31ccc18
     - name: driver-image-535
-      image: nvcr.io/nvidia/driver@sha256:6b75c7534efa6ec480e8eeea625949cd74330aad287239e1abf160622c4814f3
+      image: nvcr.io/nvidia/driver@sha256:a64d182b9d8f024ee747710475867f938386f130db8e61227e6955343e815258
     - name: driver-image-560
       image: nvcr.io/nvidia/driver@sha256:38b66a8d44cab9e2c62da9e101f32cd9dbcb5e02d8e57b47671284d374ca3695
     - name: driver-image-565
@@ -864,15 +864,15 @@ spec:
                   - name: "CONTAINER_TOOLKIT_IMAGE"
                     value: "nvcr.io/nvidia/k8s/container-toolkit@sha256:f9d5652cba9ff27f9c4d17f5d09b4a5e4aa631b5cabb9cedf4d3e8c43d6847a4"
                   - name: "DCGM_IMAGE"
-                    value: "nvcr.io/nvidia/cloud-native/dcgm@sha256:7b0ebd6c40a11b6484dc4385605372511e4e93132a44d2a3d6ec2e36c24e6783"
+                    value: "nvcr.io/nvidia/cloud-native/dcgm@sha256:e8398b2451996e5c64fc3855e7a3c93b667a85f69c0e259049d3b0012c21545d"
                   - name: "DCGM_EXPORTER_IMAGE"
-                    value: "nvcr.io/nvidia/k8s/dcgm-exporter@sha256:10ff95e83bc137796f5be76278a6b38fd31c7360e62c7e72011b428f8848c791"
+                    value: "nvcr.io/nvidia/k8s/dcgm-exporter@sha256:857f7669ae5ce2a43a3b5691fa970085098321ef8e90a896fa8dda3dee5b1d2b"
                   - name: "DEVICE_PLUGIN_IMAGE"
                     value: "nvcr.io/nvidia/k8s-device-plugin@sha256:7089559ce6153018806857f5049085bae15b3bf6f1c8bd19d8b12f707d087dea"
                   - name: "DRIVER_IMAGE"
-                    value: "nvcr.io/nvidia/driver@sha256:8a9a9e9470f64d340a7f3106a03e9622fa98e25368fbfb7ce9c416ad98f6d951"
+                    value: "nvcr.io/nvidia/driver@sha256:78e00fc8aa72f5f4925d54b4f9fbb725ca2168e890d90eadd6a497b7f31ccc18"
                   - name: "DRIVER_IMAGE-535"
-                    value: "nvcr.io/nvidia/driver@sha256:6b75c7534efa6ec480e8eeea625949cd74330aad287239e1abf160622c4814f3"
+                    value: "nvcr.io/nvidia/driver@sha256:a64d182b9d8f024ee747710475867f938386f130db8e61227e6955343e815258"
                   - name: "DRIVER_IMAGE-560"
                     value: "nvcr.io/nvidia/driver@sha256:38b66a8d44cab9e2c62da9e101f32cd9dbcb5e02d8e57b47671284d374ca3695"
                   - name: "DRIVER_IMAGE-565"

diff --git a/config/samples/nvidia_v1alpha1_nvidiadriver.yaml b/config/samples/nvidia_v1alpha1_nvidiadriver.yaml
@@ -8,7 +8,7 @@ spec:
   driverType: gpu
   repository: nvcr.io/nvidia
   image: driver
-  version: "550.127.05"
+  version: "550.127.08"
   imagePullPolicy: IfNotPresent
   imagePullSecrets: []
   nodeSelector: {}

diff --git a/deployments/gpu-operator/values.yaml b/deployments/gpu-operator/values.yaml
@@ -143,7 +143,7 @@ driver:
   usePrecompiled: false
   repository: nvcr.io/nvidia
   image: driver
-  version: "550.127.05"
+  version: "550.127.08"
   imagePullPolicy: IfNotPresent
   imagePullSecrets: []
   startupProbe:
@@ -300,7 +300,7 @@ dcgm:
   enabled: false
   repository: nvcr.io/nvidia/cloud-native
   image: dcgm
-  version: 3.3.8-1-ubuntu22.04
+  version: 3.3.9-1-ubuntu22.04
   imagePullPolicy: IfNotPresent
   args: []
   env: []
@@ -310,7 +310,7 @@ dcgmExporter:
   enabled: true
   repository: nvcr.io/nvidia/k8s
   image: dcgm-exporter
-  version: 3.3.8-3.6.0-ubuntu22.04
+  version: 3.3.9-3.6.1-ubuntu22.04
   imagePullPolicy: IfNotPresent
   env:
     - name: DCGM_EXPORTER_LISTEN

diff --git a/tests/holodeck.yaml b/tests/holodeck.yaml
@@ -12,7 +12,15 @@ spec:
     type: g4dn.xlarge
     region: us-west-1
     ingressIpRanges:
-      - 0.0.0.0/0
+    - 18.190.12.32/32
+    - 3.143.46.93/32
+    - 52.15.119.136/32
+    - 35.155.108.162/32
+    - 35.162.190.51/32
+    - 54.201.61.24/32
+    - 52.24.205.48/32
+    - 44.235.4.62/32
+    - 44.230.241.223/32
     image:
       architecture: amd64
       imageId: ami-0ce2cb35386fc22e9