Skip to content

Should give precedence to user provided env var on Universal NIM Service #577

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
visheshtanksale merged 2 commits into from
Jul 15, 2025
Merged

Should give precedence to user provided env var on Universal NIM Service #577

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6510f16
Should give precedence to user provided env var on Universal NIM Service
visheshtanksale Jul 14, 2025
6bfbd66
Addressing review comments
visheshtanksale Jul 14, 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
38 changes: 19 additions & 19 deletions api/apps/v1alpha1/nimservice_types.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -309,95 +309,95 @@ func (n *NIMService) GetLWSLeaderEnv() []corev1.EnvVar {
mpiTimeout = n.Spec.MultiNode.MPI.MPIStartTimeout
}

env = append(env,
corev1.EnvVar{
env = utils.MergeEnvVars([]corev1.EnvVar{
{
Name: "NIM_LEADER_ROLE",
Value: "1",
},
corev1.EnvVar{
{
Name: "NIM_MPI_ALLOW_RUN_AS_ROOT",
Value: "0",
},
corev1.EnvVar{
{
Name: "OMPI_MCA_orte_keep_fqdn_hostnames",
Value: "true",
},
corev1.EnvVar{
{
Name: "OMPI_MCA_plm_rsh_args",
Value: "-o ConnectionAttempts=20",
},
corev1.EnvVar{
{
Name: "NIM_NUM_COMPUTE_NODES",
Value: fmt.Sprintf("%d", n.Spec.MultiNode.Size),
},
corev1.EnvVar{
{
Name: "GPUS_PER_NODE",
Value: fmt.Sprintf("%d", n.Spec.MultiNode.GPUSPerPod),
},
corev1.EnvVar{
{
Name: "CLUSTER_START_TIMEOUT",
Value: fmt.Sprintf("%d", mpiTimeout),
},
corev1.EnvVar{
{
Name: "CLUSTER_SIZE",
ValueFrom: &corev1.EnvVarSource{
FieldRef: &corev1.ObjectFieldSelector{
FieldPath: "metadata.annotations['leaderworkerset.sigs.k8s.io/size']",
},
},
},
corev1.EnvVar{
{
Name: "GROUP_INDEX",
ValueFrom: &corev1.EnvVarSource{
FieldRef: &corev1.ObjectFieldSelector{
FieldPath: "metadata.labels['leaderworkerset.sigs.k8s.io/group-index']",
},
},
},
)
}, env)
return env
}

func (n *NIMService) GetLWSWorkerEnv() []corev1.EnvVar {
env := n.GetEnv()
env = append(env,
corev1.EnvVar{
env = utils.MergeEnvVars([]corev1.EnvVar{
{
Name: "NIM_LEADER_ROLE",
Value: "0",
},
corev1.EnvVar{
{
Name: "NIM_MPI_ALLOW_RUN_AS_ROOT",
Value: "0",
},
corev1.EnvVar{
{
Name: "NIM_NUM_COMPUTE_NODES",
Value: fmt.Sprintf("%d", n.Spec.MultiNode.Size),
},
corev1.EnvVar{
{
Name: "LEADER_NAME",
ValueFrom: &corev1.EnvVarSource{
FieldRef: &corev1.ObjectFieldSelector{
FieldPath: "metadata.annotations['leaderworkerset.sigs.k8s.io/leader-name']",
},
},
},
corev1.EnvVar{
{
Name: "NAMESPACE",
ValueFrom: &corev1.EnvVarSource{
FieldRef: &corev1.ObjectFieldSelector{
FieldPath: "metadata.namespace",
},
},
},
corev1.EnvVar{
{
Name: "LWS_NAME",
ValueFrom: &corev1.EnvVarSource{
FieldRef: &corev1.ObjectFieldSelector{
FieldPath: "metadata.labels['leaderworkerset.sigs.k8s.io/name']",
},
},
},
)
}, env)
return env
}

Expand Down
24 changes: 12 additions & 12 deletions internal/controller/platform/standalone/nimservice.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -324,7 +324,7 @@ func (r *NIMServiceReconciler) reconcileNIMService(ctx context.Context, nimServi
return ctrl.Result{}, err
}

var profileEnv *corev1.EnvVar
var profileEnv *[]corev1.EnvVar
var profile *appsv1alpha1.NIMProfile
var gpuResources *corev1.ResourceRequirements
var initContainers []corev1.Container
Expand All @@ -333,10 +333,10 @@ func (r *NIMServiceReconciler) reconcileNIMService(ctx context.Context, nimServi
var renderObj client.Object

if modelProfile != "" {
profileEnv = &corev1.EnvVar{
profileEnv = &[]corev1.EnvVar{{
Name: "NIM_MODEL_PROFILE",
Value: modelProfile,
}
}}

// Only assign GPU resources if the NIMCache is for optimized NIM
if nimCache.IsOptimizedNIM() {
Expand Down Expand Up @@ -370,20 +370,20 @@ func (r *NIMServiceReconciler) reconcileNIMService(ctx context.Context, nimServi
lwsParams.LeaderVolumes = nimService.GetLeaderVolumes(*modelPVC)
lwsParams.WorkerVolumes = nimService.GetWorkerVolumes(*modelPVC)
if nimCache.IsUniversalNIM() {
lwsParams.WorkerEnvs = append(lwsParams.WorkerEnvs, corev1.EnvVar{
lwsParams.WorkerEnvs = utils.MergeEnvVars([]corev1.EnvVar{{
Name: "NIM_MODEL_NAME",
Value: utils.DefaultModelStorePath,
})
lwsParams.LeaderEnvs = append(lwsParams.LeaderEnvs, corev1.EnvVar{
}}, lwsParams.WorkerEnvs)
lwsParams.LeaderEnvs = utils.MergeEnvVars([]corev1.EnvVar{{
Name: "NIM_MODEL_NAME",
Value: utils.DefaultModelStorePath,
})
}}, lwsParams.LeaderEnvs)
}
lwsParams.LeaderVolumeMounts = nimService.GetLeaderVolumeMounts(*modelPVC)
lwsParams.WorkerVolumeMounts = nimService.GetWorkerVolumeMounts(*modelPVC)
if profileEnv != nil {
lwsParams.WorkerEnvs = append(lwsParams.WorkerEnvs, *profileEnv)
lwsParams.LeaderEnvs = append(lwsParams.LeaderEnvs, *profileEnv)
lwsParams.WorkerEnvs = utils.MergeEnvVars(*profileEnv, lwsParams.WorkerEnvs)
lwsParams.LeaderEnvs = utils.MergeEnvVars(*profileEnv, lwsParams.LeaderEnvs)
}
if gpuResources != nil {
lwsParams.Resources = gpuResources
Expand Down Expand Up @@ -415,16 +415,16 @@ func (r *NIMServiceReconciler) reconcileNIMService(ctx context.Context, nimServi
deploymentParams.OrchestratorType = string(r.GetOrchestratorType())
deploymentParams.PodResourceClaims = shared.GetPodResourceClaims(namedDraResources)
if nimCache.IsUniversalNIM() {
deploymentParams.Env = append(deploymentParams.Env, corev1.EnvVar{
deploymentParams.Env = utils.MergeEnvVars([]corev1.EnvVar{{
Name: "NIM_MODEL_NAME",
Value: utils.DefaultModelStorePath,
})
}}, deploymentParams.Env)
}
// Setup volume mounts with model store
deploymentParams.Volumes = nimService.GetVolumes(*modelPVC)
deploymentParams.VolumeMounts = nimService.GetVolumeMounts(*modelPVC)
if profileEnv != nil {
deploymentParams.Env = append(deploymentParams.Env, *profileEnv)
deploymentParams.Env = utils.MergeEnvVars(*profileEnv, deploymentParams.Env)
}
// Auto assign GPU resources in case of the optimized profile
if gpuResources != nil {
Expand Down
Loading
Loading