[Testing] Fix NVIDIA driver issue on focal, jammy and rl9 (#1822)

GoogleCloudPlatform · Nov 6, 2024 · a0a01df · a0a01df
1 parent 6637ca8
commit a0a01df
Show file tree

Hide file tree

Showing 6 changed files with 18 additions and 12 deletions.
diff --git a/integration_test/third_party_apps_test/applications/dcgm/centos_rhel/install b/integration_test/third_party_apps_test/applications/dcgm/centos_rhel/install
@@ -51,7 +51,8 @@ setup_repo() {
 install_cuda_from_package_manager() {
     setup_repo
     install_driver_package
-    sudo yum -y install cuda
+    # TODO(b/377558109): remove the temporary fix once the repo is updated
+    sudo yum -y install cuda-toolkit cuda-demo* 
     verify_driver
 }
 

diff --git a/integration_test/third_party_apps_test/applications/dcgm/debian_ubuntu/install b/integration_test/third_party_apps_test/applications/dcgm/debian_ubuntu/install
@@ -11,8 +11,8 @@ sudo apt install -y linux-headers-${KERNEL_VERSION} software-properties-common p
 # manager's package is not working or not compitible with the GPU model
 DISTRIBUTION=$(echo $ID$VERSION_ID | sed -e 's/\.//g')
 # Need to add the keyring for installing CUDA and DCGM
-wget --no-verbose https://developer.download.nvidia.com/compute/cuda/repos/${DISTRIBUTION}/x86_64/cuda-keyring_1.0-1_all.deb
-sudo dpkg -i cuda-keyring_1.0-1_all.deb
+wget --no-verbose https://developer.download.nvidia.com/compute/cuda/repos/${DISTRIBUTION}/x86_64/cuda-keyring_1.1-1_all.deb
+sudo dpkg -i cuda-keyring_1.1-1_all.deb
 echo "Installing latest version of NVIDIA CUDA and driver"
 if [[ $ID == debian ]]; then
     sudo add-apt-repository contrib
@@ -34,7 +34,8 @@ case $DEVICE_CODE in
         ;;
     *)
         # For newer GPUs, install the latest version
-        sudo apt -y install cuda
+        # TODO(b/377558109): remove the temporary fix once the repo is updated
+        sudo apt -y install cuda-12-6
         ;;
 esac
 

diff --git a/integration_test/third_party_apps_test/applications/dcgmv1/centos_rhel/install b/integration_test/third_party_apps_test/applications/dcgmv1/centos_rhel/install
@@ -51,7 +51,8 @@ setup_repo() {
 install_cuda_from_package_manager() {
     setup_repo
     install_driver_package
-    sudo yum -y install cuda
+    # TODO(b/377558109): remove the temporary fix once the repo is updated
+    sudo yum -y install cuda-toolkit cuda-demo*
     verify_driver
 }
 

diff --git a/integration_test/third_party_apps_test/applications/dcgmv1/debian_ubuntu/install b/integration_test/third_party_apps_test/applications/dcgmv1/debian_ubuntu/install
@@ -11,8 +11,8 @@ sudo apt install -y linux-headers-${KERNEL_VERSION} software-properties-common p
 # manager's package is not working or not compitible with the GPU model
 DISTRIBUTION=$(echo $ID$VERSION_ID | sed -e 's/\.//g')
 # Need to add the keyring for installing CUDA and DCGM
-wget --no-verbose https://developer.download.nvidia.com/compute/cuda/repos/${DISTRIBUTION}/x86_64/cuda-keyring_1.0-1_all.deb
-sudo dpkg -i cuda-keyring_1.0-1_all.deb
+wget --no-verbose https://developer.download.nvidia.com/compute/cuda/repos/${DISTRIBUTION}/x86_64/cuda-keyring_1.1-1_all.deb
+sudo dpkg -i cuda-keyring_1.1-1_all.deb
 echo "Installing latest version of NVIDIA CUDA and driver"
 if [[ $ID == debian ]]; then
     sudo add-apt-repository contrib
@@ -34,7 +34,8 @@ case $DEVICE_CODE in
         ;;
     *)
         # For newer GPUs, install the latest version
-        sudo apt -y install cuda
+        # TODO(b/377558109): remove the temporary fix once the repo is updated
+        sudo apt -y install cuda-12-6
         ;;
 esac
 

diff --git a/integration_test/third_party_apps_test/applications/nvml/centos_rhel/install b/integration_test/third_party_apps_test/applications/nvml/centos_rhel/install
@@ -51,7 +51,8 @@ setup_repo() {
 install_cuda_from_package_manager() {
     setup_repo
     install_driver_package
-    sudo yum -y install cuda
+    # TODO(b/377558109): remove the temporary fix once the repo is updated
+    sudo yum -y install cuda-toolkit cuda-demo*
     verify_driver
 }
 

diff --git a/integration_test/third_party_apps_test/applications/nvml/debian_ubuntu/install b/integration_test/third_party_apps_test/applications/nvml/debian_ubuntu/install
@@ -12,8 +12,8 @@ sudo apt install -y linux-headers-${KERNEL_VERSION} software-properties-common p
 # manager's package is not working or not compitible with the GPU model
 DISTRIBUTION=$(echo $ID$VERSION_ID | sed -e 's/\.//g')
 echo "Installing latest version of NVIDIA CUDA and driver"
-wget --no-verbose https://developer.download.nvidia.com/compute/cuda/repos/${DISTRIBUTION}/x86_64/cuda-keyring_1.0-1_all.deb
-sudo dpkg -i cuda-keyring_1.0-1_all.deb
+wget --no-verbose https://developer.download.nvidia.com/compute/cuda/repos/${DISTRIBUTION}/x86_64/cuda-keyring_1.1-1_all.deb
+sudo dpkg -i cuda-keyring_1.1-1_all.deb
 if [[ $ID == debian ]]; then
     sudo add-apt-repository contrib
 fi
@@ -34,7 +34,8 @@ case $DEVICE_CODE in
         ;;
     *)
         # For newer GPUs, install the latest version
-        sudo apt -y install cuda
+        # TODO(b/377558109): remove the temporary fix once the repo is updated 
+        sudo apt -y install cuda-12-6
         ;;
 esac