Skip to content

[enterprise-4.19] OSDOCS 16489 Add GPU usage with CMA docs -- NEEDED for 4.20! #100572

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
merged 1 commit into from
Oct 15, 2025
Merged

[enterprise-4.19] OSDOCS 16489 Add GPU usage with CMA docs -- NEEDED for 4.20! #100572

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
40 changes: 40 additions & 0 deletions modules/nodes-cma-autoscaling-custom-trigger-prom-gpu.adoc
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,40 @@
// Module included in the following assemblies:
//
// * nodes/cma/nodes-cma-autoscaling-custom-trigger.adoc

:_mod-docs-content-type: CONCEPT
[id="nodes-cma-autoscaling-custom-trigger-prom-gpu_{context}"]
= Configuring GPU-based autoscaling with Prometheus and DCGM metrics

You can use the Custom Metrics Autoscaler with NVIDIA Data Center GPU Manager (DCGM) metrics to scale workloads based on GPU utilization. This is particularly useful for AI and machine learning workloads that require GPU resources.

.Example scaled object with a Prometheus target for GPU-based autoscaling
[source,yaml,options="nowrap"]
----
apiVersion: keda.sh/v1alpha1
kind: ScaledObject
metadata:
name: gpu-scaledobject
namespace: my-namespace
spec:
scaleTargetRef:
kind: Deployment
name: gpu-deployment
minReplicaCount: 1 <1>
maxReplicaCount: 5 <2>
triggers:
- type: prometheus
metadata:
serverAddress: https://thanos-querier.openshift-monitoring.svc.cluster.local:9092
namespace: my-namespace
metricName: gpu_utilization
threshold: '90' <3>
query: SUM(DCGM_FI_DEV_GPU_UTIL{instance=~".+", gpu=~".+"}) <4>
authModes: bearer
authenticationRef:
name: keda-trigger-auth-prometheus
----
<1> Specifies the minimum number of replicas to maintain. For GPU workloads, this should not be set to `0` to ensure that metrics continue to be collected.
<2> Specifies the maximum number of replicas allowed during scale-up operations.
<3> Specifies the GPU utilization percentage threshold that triggers scaling. When the average GPU utilization exceeds 90%, the autoscaler scales up the deployment.
<4> Specifies a Prometheus query using NVIDIA DCGM metrics to monitor GPU utilization across all GPU devices. The `DCGM_FI_DEV_GPU_UTIL` metric provides GPU utilization percentages.
1 change: 1 addition & 0 deletions nodes/cma/nodes-cma-autoscaling-custom-trigger.adoc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -22,6 +22,7 @@ You can configure a certificate authority xref:../../nodes/cma/nodes-cma-autosca
// assemblies.

include::modules/nodes-cma-autoscaling-custom-trigger-prom.adoc[leveloffset=+1]
include::modules/nodes-cma-autoscaling-custom-trigger-prom-gpu.adoc[leveloffset=+2]
include::modules/nodes-cma-autoscaling-custom-prometheus-config.adoc[leveloffset=+2]

[role="_additional-resources"]
Expand Down