Merge pull request #280 from ArangoGutierrez/nv_driver_version

Make NVDRIVER configurable from env file

Author: ArangoGutierrez

README.md

@@ -113,4 +113,59 @@ Dryrun environment holodeck 🔍
 ✔       Checking if image ami-0fe8bec493a81c7da is supported in region eu-north-1
 ✔      Resolving dependencies 📦
 Dryrun succeeded 🎉
-```
+```
+
+## Supported Cuda-Drivers
+
+Supported Nvidia drivers are: 
+
+```yaml
+  nvidiaDriver:
+    install: true
+    version: <version>
+```
+Where `<version>` can be a prefix of any package version. The following are example package versions:
+
+- 570.86.15-0ubuntu1
+- 570.86.10-0ubuntu1
+- 565.57.01-0ubuntu1
+- 560.35.05-0ubuntu1
+- 560.35.03-1
+- 560.28.03-1
+- 555.42.06-1
+- 555.42.02-1
+- 550.144.03-0ubuntu1
+- 550.127.08-0ubuntu1
+- 550.127.05-0ubuntu1
+- 550.90.12-0ubuntu1
+- 550.90.07-1
+- 550.54.15-1
+- 550.54.14-1
+- 545.23.08-1
+- 545.23.06-1
+- 535.230.02-0ubuntu1
+- 535.216.03-0ubuntu1
+- 535.216.01-0ubuntu1
+- 535.183.06-1
+- 535.183.01-1
+- 535.161.08-1
+- 535.161.07-1
+- 535.154.05-1
+- 535.129.03-1
+- 535.104.12-1
+- 535.104.05-1
+- 535.86.10-1
+- 535.54.03-1
+- 530.30.02-1
+- 525.147.05-1
+- 525.125.06-1
+- 525.105.17-1
+- 525.85.12-1
+- 525.60.13-1
+- 520.61.05-1
+- 515.105.01-1
+- 515.86.01-1
+- 515.65.07-1
+- 515.65.01-1
+- 515.48.07-1
+- 515.43.04-1

pkg/provisioner/dependency.go

@@ -50,7 +50,7 @@ var (
 type ProvisionFunc func(tpl *bytes.Buffer, env v1alpha1.Environment) error
 
 func nvdriver(tpl *bytes.Buffer, env v1alpha1.Environment) error {
-	nvdriver := templates.NewNvDriver()
+	nvdriver := templates.NewNvDriver(env)
 	return nvdriver.Execute(tpl, env)
 }
 

pkg/provisioner/templates/common.go

@@ -38,7 +38,7 @@ install_packages_with_retry() {
         echo "Attempt $i to install packages: ${packages[@]}"
         
         # Attempt to install packages
-        sudo apt-get install -y "${packages[@]}"
+        sudo apt-get install -y --no-install-recommends "${packages[@]}"
 
         # Check if the last command failed and the error is related to unsigned repository
         if [ $? -ne 0 ] && grep -q 'NO_PUBKEY' <<< "$(tail -n 1 /var/lib/dpkg/status 2>/dev/null)"; then

pkg/provisioner/templates/nv-driver.go

@@ -34,21 +34,25 @@ wget https://developer.download.nvidia.com/compute/cuda/repos/$distribution/x86_
 sudo dpkg -i cuda-keyring_1.1-1_all.deb
 
 with_retry 3 10s sudo apt-get update
-install_packages_with_retry cuda-drivers
+install_packages_with_retry cuda-drivers{{if .Version}}={{.Version}}{{end}}
 
-nvidia-smi -L
+nvidia-smi
 `
 
 type NvDriver struct {
+	// Version -- if specified -- indicates the version of the `cuda-drivers` package to install.
+	Version string
 }
 
-func NewNvDriver() *NvDriver {
-	return &NvDriver{}
+func NewNvDriver(env v1alpha1.Environment) *NvDriver {
+	return &NvDriver{
+		Version: env.Spec.NVIDIADriver.Version,
+	}
 }
 
 func (t *NvDriver) Execute(tpl *bytes.Buffer, env v1alpha1.Environment) error {
 	nvDriverTemplate := template.Must(template.New("nv-driver").Parse(NvDriverTemplate))
-	err := nvDriverTemplate.Execute(tpl, &NvDriver{})
+	err := nvDriverTemplate.Execute(tpl, t)
 	if err != nil {
 		return fmt.Errorf("failed to execute nv-driver template: %v", err)
 	}