Skip to content

fix: mpi flow and add resourceClaim #3446

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
nvrohanv merged 12 commits into from
Oct 23, 2025
Merged

fix: mpi flow and add resourceClaim #3446

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
12 commits
Select commit Hold shift + click to select a range
56cb033
fix mpi flow and add resourceClaim
nvrohanv Oct 6, 2025
93c46f9
Merge branch 'main' into nvrohanv/fix-mpi-flow
nvrohanv Oct 6, 2025
bc6aa02
Merge branch 'main' into nvrohanv/fix-mpi-flow
nvrohanv Oct 7, 2025
b2f1e33
Merge branch 'main' into nvrohanv/fix-mpi-flow
nvrohanv Oct 7, 2025
98ae215
Merge branch 'main' into nvrohanv/fix-mpi-flow
nvrohanv Oct 20, 2025
c457fd5
Merge branch 'main' into nvrohanv/fix-mpi-flow
nvrohanv Oct 20, 2025
2c6a423
Merge branch 'main' into nvrohanv/fix-mpi-flow
nvrohanv Oct 20, 2025
f50d8ae
Merge branch 'main' into nvrohanv/fix-mpi-flow
nvrohanv Oct 21, 2025
564a39f
Merge branch 'main' into nvrohanv/fix-mpi-flow
nvrohanv Oct 21, 2025
ffa22e8
Merge branch 'main' into nvrohanv/fix-mpi-flow
nvrohanv Oct 22, 2025
2032793
Merge branch 'main' into nvrohanv/fix-mpi-flow
nvrohanv Oct 22, 2025
5c5e944
Merge branch 'main' into nvrohanv/fix-mpi-flow
nvrohanv Oct 23, 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
20 changes: 20 additions & 0 deletions deploy/cloud/helm/crds/templates/nvidia.com_dynamocomponentdeployments.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10181,6 +10181,26 @@ spec:
Resources requested and limits for this component, including CPU, memory,
GPUs/devices, and any runtime-specific resources.
properties:
claims:
items:
description: ResourceClaim references one entry in PodSpec.ResourceClaims.
properties:
name:
description: |-
Name must match the name of one entry in pod.spec.resourceClaims of
the Pod where this field is used. It makes that resource available
inside a container.
type: string
request:
description: |-
Request is the name chosen for a request in the referenced claim.
If empty, everything from the claim is made available, otherwise
only the result of this request.
type: string
required:
- name
type: object
type: array
limits:
properties:
cpu:
Expand Down
20 changes: 20 additions & 0 deletions deploy/cloud/helm/crds/templates/nvidia.com_dynamographdeployments.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10312,6 +10312,26 @@ spec:
Resources requested and limits for this component, including CPU, memory,
GPUs/devices, and any runtime-specific resources.
properties:
claims:
items:
description: ResourceClaim references one entry in PodSpec.ResourceClaims.
properties:
name:
description: |-
Name must match the name of one entry in pod.spec.resourceClaims of
the Pod where this field is used. It makes that resource available
inside a container.
type: string
request:
description: |-
Request is the name chosen for a request in the referenced claim.
If empty, everything from the claim is made available, otherwise
only the result of this request.
type: string
required:
- name
type: object
type: array
limits:
properties:
cpu:
Expand Down
5 changes: 3 additions & 2 deletions deploy/cloud/operator/api/dynamo/common/common.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,8 +32,9 @@ type ResourceItem struct {
}

type Resources struct {
Requests *ResourceItem `json:"requests,omitempty"`
Limits *ResourceItem `json:"limits,omitempty"`
Requests *ResourceItem `json:"requests,omitempty"`
Limits *ResourceItem `json:"limits,omitempty"`
Claims []corev1.ResourceClaim `json:"claims,omitempty"`
}

type DeploymentTargetHPAConf struct {
Expand Down
5 changes: 5 additions & 0 deletions deploy/cloud/operator/api/dynamo/common/zz_generated.deepcopy.go
View file
Open in desktop

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

20 changes: 20 additions & 0 deletions deploy/cloud/operator/config/crd/bases/nvidia.com_dynamocomponentdeployments.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10181,6 +10181,26 @@ spec:
Resources requested and limits for this component, including CPU, memory,
GPUs/devices, and any runtime-specific resources.
properties:
claims:
items:
description: ResourceClaim references one entry in PodSpec.ResourceClaims.
properties:
name:
description: |-
Name must match the name of one entry in pod.spec.resourceClaims of
the Pod where this field is used. It makes that resource available
inside a container.
type: string
request:
description: |-
Request is the name chosen for a request in the referenced claim.
If empty, everything from the claim is made available, otherwise
only the result of this request.
type: string
required:
- name
type: object
type: array
limits:
properties:
cpu:
Expand Down
20 changes: 20 additions & 0 deletions deploy/cloud/operator/config/crd/bases/nvidia.com_dynamographdeployments.yaml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10312,6 +10312,26 @@ spec:
Resources requested and limits for this component, including CPU, memory,
GPUs/devices, and any runtime-specific resources.
properties:
claims:
items:
description: ResourceClaim references one entry in PodSpec.ResourceClaims.
properties:
name:
description: |-
Name must match the name of one entry in pod.spec.resourceClaims of
the Pod where this field is used. It makes that resource available
inside a container.
type: string
request:
description: |-
Request is the name chosen for a request in the referenced claim.
If empty, everything from the claim is made available, otherwise
only the result of this request.
type: string
required:
- name
type: object
type: array
limits:
properties:
cpu:
Expand Down
6 changes: 6 additions & 0 deletions deploy/cloud/operator/internal/controller_common/resource.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -468,6 +468,12 @@ func GetResourcesConfig(resources *common.Resources) (*corev1.ResourceRequiremen
currentResources.Requests[corev1.ResourceName(k)] = q
}
}
if resources.Claims != nil {
if currentResources.Claims == nil {
currentResources.Claims = make([]corev1.ResourceClaim, 0)
}
currentResources.Claims = append(currentResources.Claims, resources.Claims...)
}
return currentResources, nil
}

Expand Down
4 changes: 2 additions & 2 deletions deploy/cloud/operator/internal/dynamo/backend_trtllm.go
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -143,12 +143,12 @@ func (b *TRTLLMBackend) setupLeaderContainer(container *corev1.Container, number

// Build mpirun command with explicit SSH configuration and environment variables
// Wrap the entire command (trtllm-llmapi-launch + original command) in bash -c for proper shell interpretation
wrappedCommand := fmt.Sprintf("bash -c 'source /opt/dynamo/venv/bin/activate && trtllm-llmapi-launch %s'", originalCommand)
wrappedCommand := fmt.Sprintf("bash -c 'trtllm-llmapi-launch %s'", originalCommand)

// Generate environment variable flags for mpirun
envVarsStr := generateEnvVarFlags(container.Env)

mpirunCmd := fmt.Sprintf("mpirun --oversubscribe -n %d -H %s --mca pml ob1 --mca plm_rsh_args \"-p %d -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" %s %s",
mpirunCmd := fmt.Sprintf("mpirun --allow-run-as-root --oversubscribe -n %d -H %s --mca pml ob1 --mca plm_rsh_args \"-p %d -o StrictHostKeyChecking=no -i ~/.ssh/id_rsa\" %s %s",
totalGPUs,
workerHosts,
commonconsts.MpiRunSshPort,
Expand Down
Loading
Loading