Add nvidia-fabricmanager to dstack VM images

dstackai · Jul 2, 2024 · adf7c8d · adf7c8d
1 parent 88d0936
commit adf7c8d
Show file tree

Hide file tree

Showing 14 changed files with 23 additions and 17 deletions.
diff --git a/.github/workflows/docker.yml b/.github/workflows/docker.yml
@@ -224,12 +224,14 @@ jobs:
         run: |
           pip install .[oci]
       - name: Copy image to target regions
+        if: ${{ !inputs.staging }}
         run: |
           python scripts/oci_image_tools.py copy \
             --image ${BUILD_PREFIX}dstack${{ matrix.variant }}-${{ inputs.image_version }} \
             --from $OCI_REGION \
             --compartment $OCI_COMPARTMENT
       - name: Publish image in OCI Marketplace
+        if: ${{ !inputs.staging }}
         run: |
           python scripts/oci_image_tools.py publish \
             --image ${BUILD_PREFIX}dstack${{ matrix.variant }}-${{ inputs.image_version }} \

diff --git a/scripts/packer/aws-image-cuda.json b/scripts/packer/aws-image-cuda.json
@@ -7,8 +7,8 @@
     "base_ami": "ami-0cffefff2d52e0a23",
     "instance_type": "c5.large",
     "subnet_id": "subnet-c39cb6a5",
-    "docker_version": "20.10.17",
-    "cuda_drivers_version": "535.54.03-1",
+    "docker_version": "",
+    "cuda_drivers_version": "",
     "build_prefix": "",
     "ami_regions": "",
     "ami_groups": "",

diff --git a/scripts/packer/aws-image.json b/scripts/packer/aws-image.json
@@ -7,7 +7,7 @@
     "base_ami": "ami-0cffefff2d52e0a23",
     "instance_type": "c5.large",
     "subnet_id": "subnet-c39cb6a5",
-    "docker_version": "20.10.17",
+    "docker_version": "",
     "build_prefix": "",
     "ami_regions": "",
     "ami_groups": "",

diff --git a/scripts/packer/azure-image-cuda.json b/scripts/packer/azure-image-cuda.json
@@ -8,8 +8,8 @@
     "azure_location": "westeurope",
     "azure_vm_size": "Standard_DS1_v2",
     "build_prefix": "",
-    "docker_version": "20.10.17",
-    "cuda_drivers_version": "535.54.03-1",
+    "docker_version": "",
+    "cuda_drivers_version": "",
     "image_version": ""
   },
   "builders": [{

diff --git a/scripts/packer/azure-image-grid.json b/scripts/packer/azure-image-grid.json
@@ -8,7 +8,7 @@
     "azure_location": "westeurope",
     "azure_vm_size": "Standard_DS1_v2",
     "build_prefix": "",
-    "docker_version": "20.10.17",
+    "docker_version": "",
     "image_version": ""
   },
   "builders": [{

diff --git a/scripts/packer/azure-image.json b/scripts/packer/azure-image.json
@@ -8,7 +8,7 @@
     "azure_location": "westeurope",
     "azure_vm_size": "Standard_DS1_v2",
     "build_prefix": "",
-    "docker_version": "20.10.17",
+    "docker_version": "",
     "image_version": ""
   },
   "builders": [{

diff --git a/scripts/packer/gcp-image-cuda.json b/scripts/packer/gcp-image-cuda.json
@@ -1,8 +1,8 @@
 {
   "variables": {
     "build_prefix": "",
-    "docker_version": "20.10.17",
-    "cuda_drivers_version": "535.54.03-1",
+    "docker_version": "",
+    "cuda_drivers_version": "",
     "image_version": ""
   },
   "builders": [

diff --git a/scripts/packer/gcp-image.json b/scripts/packer/gcp-image.json
@@ -1,7 +1,7 @@
 {
   "variables": {
     "build_prefix": "",
-    "docker_version": "20.10.17",
+    "docker_version": "",
     "image_version": ""
   },
   "builders": [

diff --git a/scripts/packer/locals.pkr.hcl b/scripts/packer/locals.pkr.hcl
@@ -1,6 +1,6 @@
 locals {
   clean_image_version = regex_replace(var.image_version, "[^a-z0-9-]", "-")
   image_name = "${var.build_prefix}dstack-${local.clean_image_version}"
-  docker_version = "20.10.17"
-  cuda_drivers_version = "535.54.03-1"
+  docker_version = ""
+  cuda_drivers_version = ""
 }
diff --git a/scripts/packer/oci-image-cuda.json b/scripts/packer/oci-image-cuda.json
@@ -1,7 +1,8 @@
 {
   "variables": {
     "build_prefix": "",
-    "docker_version": "20.10.17",
+    "docker_version": "",
+    "cuda_drivers_version": "",
     "image_version": "",
     "oci_availability_domain": "",
     "oci_compartment_ocid": "",

diff --git a/scripts/packer/oci-image.json b/scripts/packer/oci-image.json
@@ -1,7 +1,7 @@
 {
   "variables": {
     "build_prefix": "",
-    "docker_version": "20.10.17",
+    "docker_version": "",
     "image_version": "",
     "oci_availability_domain": "",
     "oci_compartment_ocid": "",

diff --git a/scripts/packer/provisioners/cuda.sh b/scripts/packer/provisioners/cuda.sh
@@ -16,4 +16,7 @@ rm cuda-keyring_1.0-1_all.deb
 
 CUDA_BRANCH=$(cut -d '.' -f 1 <<< "$CUDA_DRIVERS_VERSION")
 sudo apt-get update
-sudo DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends cuda-drivers-$CUDA_BRANCH=$CUDA_DRIVERS_VERSION
+sudo DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
+    cuda-drivers-$CUDA_BRANCH=$CUDA_DRIVERS_VERSION \
+    nvidia-fabricmanager-$CUDA_BRANCH=$CUDA_DRIVERS_VERSION
+sudo systemctl enable nvidia-fabricmanager
diff --git a/scripts/packer/provisioners/install-nvidia-grid-driver-for-azure.sh b/scripts/packer/provisioners/install-nvidia-grid-driver-for-azure.sh
@@ -8,7 +8,7 @@ sudo apt-get update
 sudo DEBIAN_FRONTEND=noninteractive apt-get install build-essential linux-azure -y
 
 wget --no-verbose -O NVIDIA-Linux-x86_64-grid.run \
-    https://download.microsoft.com/download/1/4/4/14450d0e-a3f2-4b0a-9bb4-a8e729e986c4/NVIDIA-Linux-x86_64-535.154.05-grid-azure.run
+    https://download.microsoft.com/download/8/d/a/8da4fb8e-3a9b-4e6a-bc9a-72ff64d7a13c/NVIDIA-Linux-x86_64-535.161.08-grid-azure.run
 chmod +x NVIDIA-Linux-x86_64-grid.run
 sudo ./NVIDIA-Linux-x86_64-grid.run --silent --disable-nouveau
 rm NVIDIA-Linux-x86_64-grid.run

diff --git a/scripts/packer/versions.json b/scripts/packer/versions.json
@@ -1,4 +1,4 @@
 {
   "docker_version": "20.10.17",
-  "cuda_drivers_version": "535.54.03-1"
+  "cuda_drivers_version": "535.183.01-1"
 }