Skip to content

Add no-runtime-configure option to the toolkit installer #1521

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
elezar merged 4 commits into from
Jan 23, 2026
Merged

Add no-runtime-configure option to the toolkit installer #1521

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
9dc0506
[no-relnote] Quick return from installer on noDaemon
elezar Jan 14, 2026
0b0302e
[no-relnote] Move runtime name validation to runtime package
elezar Jan 14, 2026
7c9a522
Add interface for runtime configuration to installer
elezar Jan 14, 2026
3791812
Add --no-runtime-config option to toolkit installer
elezar Dec 8, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
56 changes: 47 additions & 9 deletions cmd/nvidia-ctk-installer/container/runtime/runtime.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -138,6 +138,15 @@ func Flags(opts *Options) []cli.Flag {

// Validate checks whether the specified options are valid
func (opts *Options) Validate(logger logger.Interface, c *cli.Command, runtime string, toolkitRoot string, to *toolkit.Options) error {
// Check that the specified runtime is supported.
switch runtime {
case containerd.Name:
case crio.Name:
case docker.Name:
default:
return fmt.Errorf("invalid runtime %q; expected one of [containerd | crio | docker]", runtime)
}

// We set this option here to ensure that it is available in future calls.
opts.RuntimeDir = toolkitRoot

Expand Down Expand Up @@ -212,41 +221,70 @@ func (opts *Options) Validate(logger logger.Interface, c *cli.Command, runtime s
return nil
}

func Setup(c *cli.Command, opts *Options, runtime string) error {
switch runtime {
type Configurer interface {
Cleanup(*cli.Command, *Options) error
GetLowlevelRuntimePaths(*Options) ([]string, error)
Setup(*cli.Command, *Options) error
}

type runtime string
type noopRuntimeConfigurer struct{}

// NewConfigurer is a factory method for creating a runtime configurer.
func NewConfigurer(name string, noConfigureRuntime bool, nriEnabled bool) Configurer {
if noConfigureRuntime || nriEnabled {
return &noopRuntimeConfigurer{}
}
return runtime(name)
}

func (r runtime) Setup(c *cli.Command, opts *Options) error {
switch string(r) {
case containerd.Name:
return containerd.Setup(c, &opts.Options, &opts.containerdOptions)
case crio.Name:
return crio.Setup(c, &opts.Options, &opts.crioOptions)
case docker.Name:
return docker.Setup(c, &opts.Options)
default:
return fmt.Errorf("undefined runtime %v", runtime)
return fmt.Errorf("undefined runtime %v", r)
}
}

func Cleanup(c *cli.Command, opts *Options, runtime string) error {
switch runtime {
func (r runtime) Cleanup(c *cli.Command, opts *Options) error {
switch string(r) {
case containerd.Name:
return containerd.Cleanup(c, &opts.Options, &opts.containerdOptions)
case crio.Name:
return crio.Cleanup(c, &opts.Options, &opts.crioOptions)
case docker.Name:
return docker.Cleanup(c, &opts.Options)
default:
return fmt.Errorf("undefined runtime %v", runtime)
return fmt.Errorf("undefined runtime %v", r)
}
}

func GetLowlevelRuntimePaths(opts *Options, runtime string) ([]string, error) {
switch runtime {
func (r runtime) GetLowlevelRuntimePaths(opts *Options) ([]string, error) {
switch string(r) {
case containerd.Name:
return containerd.GetLowlevelRuntimePaths(&opts.Options, &opts.containerdOptions)
case crio.Name:
return crio.GetLowlevelRuntimePaths(&opts.Options)
case docker.Name:
return docker.GetLowlevelRuntimePaths(&opts.Options)
default:
return nil, fmt.Errorf("undefined runtime %v", runtime)
return nil, fmt.Errorf("undefined runtime %v", r)
}
}

func (r noopRuntimeConfigurer) Cleanup(_ *cli.Command, _ *Options) error {
return nil
}

func (r noopRuntimeConfigurer) GetLowlevelRuntimePaths(_ *Options) ([]string, error) {
return nil, nil
}

func (r noopRuntimeConfigurer) Setup(_ *cli.Command, _ *Options) error {
return nil
}
71 changes: 40 additions & 31 deletions cmd/nvidia-ctk-installer/main.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,7 +33,6 @@ const (
defaultNRIPluginIdx uint = 10
)

var availableRuntimes = map[string]struct{}{"docker": {}, "crio": {}, "containerd": {}}
var defaultLowLevelRuntimes = []string{"runc", "crun"}

var waitingForSignal = make(chan bool, 1)
Expand All @@ -44,7 +43,6 @@ type options struct {
toolkitInstallDir string

noDaemon bool
runtime string
pidFile string
sourceRoot string
packageType string
Expand All @@ -54,7 +52,10 @@ type options struct {
nriSocket string

toolkitOptions toolkit.Options
runtimeOptions runtime.Options

noRuntimeConfig bool
runtime string
runtimeOptions runtime.Options
}

func (o options) toolkitRoot() string {
Expand Down Expand Up @@ -113,6 +114,14 @@ func (a app) build() *cli.Command {
Destination: &options.noDaemon,
Sources: cli.EnvVars("NO_DAEMON"),
},
&cli.BoolFlag{
Name: "no-runtime-config",
Usage: "Disables the configuration of a container runtime. This is used in cases where the runtime has " +
"already been configured for use with the toolkit, and the installer is only used to deploy the " +
"components of the NVIDIA Container Toolkit.",
Destination: &options.noRuntimeConfig,
Sources: cli.EnvVars("NO_RUNTIME_CONFIG"),
},
&cli.BoolFlag{
Name: "enable-nri-plugin",
Aliases: []string{"p"},
Expand Down Expand Up @@ -142,9 +151,10 @@ func (a app) build() *cli.Command {
Sources: cli.EnvVars("NRI_SOCKET"),
},
&cli.StringFlag{
Name: "runtime",
Aliases: []string{"r"},
Usage: "the runtime to setup on this node. One of {'docker', 'crio', 'containerd'}",
Name: "runtime",
Aliases: []string{"r"},
Usage: "the runtime to setup on this node. One of {'docker', 'crio', 'containerd'}. " +
"This setting is ignored if --no-runtime-config is specified.",
Value: defaultRuntime,
Destination: &options.runtime,
Sources: cli.EnvVars("RUNTIME"),
Expand Down Expand Up @@ -212,19 +222,18 @@ func (a *app) validateFlags(c *cli.Command, o *options) error {
if o.toolkitInstallDir == "" {
return fmt.Errorf("the install root must be specified")
}
if _, exists := availableRuntimes[o.runtime]; !exists {
return fmt.Errorf("unknown runtime: %v", o.runtime)
}
if filepath.Base(o.pidFile) != toolkitPidFilename {
return fmt.Errorf("invalid toolkit.pid path %v", o.pidFile)
}

if err := a.toolkit.ValidateOptions(&o.toolkitOptions); err != nil {
return err
}

if err := o.runtimeOptions.Validate(a.logger, c, o.runtime, o.toolkitRoot(), &o.toolkitOptions); err != nil {
return err
}

return nil
}

Expand All @@ -238,8 +247,10 @@ func (a *app) Run(ctx context.Context, c *cli.Command, o *options) error {
}
defer a.shutdown(o.pidFile)

runtimeConfigurer := runtime.NewConfigurer(o.runtime, o.noRuntimeConfig, o.enableNRIPlugin)

if len(o.toolkitOptions.ContainerRuntimeRuntimes) == 0 {
lowlevelRuntimePaths, err := runtime.GetLowlevelRuntimePaths(&o.runtimeOptions, o.runtime)
lowlevelRuntimePaths, err := runtimeConfigurer.GetLowlevelRuntimePaths(&o.runtimeOptions)
if err != nil {
return fmt.Errorf("unable to determine runtime options: %w", err)
}
Expand All @@ -253,33 +264,31 @@ func (a *app) Run(ctx context.Context, c *cli.Command, o *options) error {
return fmt.Errorf("unable to install toolkit: %v", err)
}

if !o.enableNRIPlugin {
err = runtime.Setup(c, &o.runtimeOptions, o.runtime)
if err != nil {
return fmt.Errorf("unable to setup runtime: %w", err)
}
err = runtimeConfigurer.Setup(c, &o.runtimeOptions)
if err != nil {
return fmt.Errorf("unable to setup runtime: %w", err)
}

if !o.noDaemon {
if o.enableNRIPlugin {
nriPlugin, err := a.startNRIPluginServer(ctx, o)
if err != nil {
a.logger.Errorf("unable to start NRI plugin server: %v", err)
}
defer nriPlugin.Stop()
}
if o.noDaemon {
return nil
}

err = a.waitForSignal()
if o.enableNRIPlugin {
nriPlugin, err := a.startNRIPluginServer(ctx, o)
if err != nil {
return fmt.Errorf("unable to wait for signal: %v", err)
a.logger.Errorf("unable to start NRI plugin server: %v", err)
}
defer nriPlugin.Stop()
}

if !o.enableNRIPlugin {
err = runtime.Cleanup(c, &o.runtimeOptions, o.runtime)
if err != nil {
return fmt.Errorf("unable to cleanup runtime: %v", err)
}
}
err = a.waitForSignal()
if err != nil {
return fmt.Errorf("unable to wait for signal: %v", err)
}

err = runtimeConfigurer.Cleanup(c, &o.runtimeOptions)
if err != nil {
return fmt.Errorf("unable to cleanup runtime: %v", err)
}

return nil
Expand Down