chore(ci): run e2e-aws-nvidia with zfs extension enabled

Run e2e-aws-nvidia-oss with zfs extension enabled. Also fix the iscsi tests to get transport info using the new disks api. Signed-off-by: Noel Georgi <git@frezbo.dev>
siderolabs · Jun 24, 2024 · 4701498 · 4701498
1 parent 86a3222
commit 4701498
Show file tree

Hide file tree

Showing 7 changed files with 9 additions and 8 deletions.
diff --git a/.github/workflows/ci.yaml b/.github/workflows/ci.yaml
@@ -1,6 +1,6 @@
 # THIS FILE WAS AUTOMATICALLY GENERATED, PLEASE DO NOT EDIT.
 #
-# Generated on 2024-06-11T15:17:44Z by kres 7360563.
+# Generated on 2024-06-24T09:42:29Z by kres 4c9f215.
 
 name: default
 concurrency:
@@ -732,7 +732,7 @@ jobs:
           PUSH: "true"
           REGISTRY: registry.dev.siderolabs.io
         run: |
-          make nvidia-container-toolkit nvidia-open-gpu-kernel-modules extensions-metadata -C _out/extensions
+          make nvidia-container-toolkit nvidia-open-gpu-kernel-modules zfs extensions-metadata -C _out/extensions
       - name: e2e-aws-prepare
         env:
           E2E_AWS_TARGET: nvidia-oss

diff --git a/.github/workflows/integration-aws-nvidia-oss-cron.yaml b/.github/workflows/integration-aws-nvidia-oss-cron.yaml
@@ -1,6 +1,6 @@
 # THIS FILE WAS AUTOMATICALLY GENERATED, PLEASE DO NOT EDIT.
 #
-# Generated on 2024-05-27T16:20:10Z by kres bcb280a.
+# Generated on 2024-06-10T14:11:50Z by kres 7360563.
 
 name: integration-aws-nvidia-oss-cron
 concurrency:
@@ -112,7 +112,7 @@ jobs:
           PUSH: "true"
           REGISTRY: registry.dev.siderolabs.io
         run: |
-          make nvidia-container-toolkit nvidia-open-gpu-kernel-modules extensions-metadata -C _out/extensions
+          make nvidia-container-toolkit nvidia-open-gpu-kernel-modules zfs extensions-metadata -C _out/extensions
       - name: e2e-aws-prepare
         env:
           E2E_AWS_TARGET: nvidia-oss

diff --git a/.kres.yaml b/.kres.yaml
@@ -1582,7 +1582,7 @@ spec:
           nonMakeStep: true
           command: cat _out/talos-metadata >> "$GITHUB_ENV"
         - name: build extensions
-          command: nvidia-container-toolkit nvidia-open-gpu-kernel-modules extensions-metadata
+          command: nvidia-container-toolkit nvidia-open-gpu-kernel-modules zfs extensions-metadata
           arguments:
             - -C
             - _out/extensions

diff --git a/hack/test/e2e-aws-prepare.sh b/hack/test/e2e-aws-prepare.sh
@@ -19,7 +19,7 @@ function get_ami_id() {
 function cloud_image_upload_with_extensions() {
   case "${1}" in
     nvidia-oss)
-      EXTENSIONS=$(jq -R < "${EXTENSIONS_METADATA_FILE}" | jq -rs 'map(select(. | contains("nvidia") and (contains("nvidia-fabricmanager") or contains("nonfree-kmod-nvidia") | not))) | .[] |= "--system-extension-image=" + . | join(" ")')
+      EXTENSIONS=$(jq -R < "${EXTENSIONS_METADATA_FILE}" | jq -rs 'map(select(. | (contains("nvidia") or contains("zfs")) and (contains("nvidia-fabricmanager") or contains("nonfree-kmod-nvidia") | not))) | .[] |= "--system-extension-image=" + . | join(" ")')
       ;;
     nvidia-oss-fabricmanager)
       EXTENSIONS=$(jq -R < "${EXTENSIONS_METADATA_FILE}" | jq -rs 'map(select(. | contains("nvidia") and (contains("nonfree-kmod-nvidia") | not))) | .[] |= "--system-extension-image=" + . | join(" ")')

diff --git a/hack/test/tfvars/nvidia.yaml b/hack/test/tfvars/nvidia.yaml
@@ -5,5 +5,6 @@ machine:
       - name: nvidia_uvm
       - name: nvidia_drm
       - name: nvidia_modeset
+      - name: zfs
   sysctls:
     net.core.bpf_jit_harden: 1
diff --git a/website/content/v1.8/talos-guides/configuration/nvidia-gpu-proprietary.md b/website/content/v1.8/talos-guides/configuration/nvidia-gpu-proprietary.md
@@ -141,7 +141,7 @@ kubectl run \
   nvidia-test \
   --restart=Never \
   -ti --rm \
-  --image nvcr.io/nvidia/cuda:12.1.0-base-ubuntu22.04 \
+  --image nvcr.io/nvidia/cuda:12.5.0-base-ubuntu22.04 \
   --overrides '{"spec": {"runtimeClassName": "nvidia"}}' \
   nvidia-smi
 ```
diff --git a/website/content/v1.8/talos-guides/configuration/nvidia-gpu.md b/website/content/v1.8/talos-guides/configuration/nvidia-gpu.md
@@ -140,7 +140,7 @@ kubectl run \
   nvidia-test \
   --restart=Never \
   -ti --rm \
-  --image nvcr.io/nvidia/cuda:12.1.0-base-ubuntu22.04 \
+  --image nvcr.io/nvidia/cuda:12.5.0-base-ubuntu22.04 \
   --overrides '{"spec": {"runtimeClassName": "nvidia"}}' \
   nvidia-smi
 ```