quorate/vendor/infrastructure/nvidia/helm-releases.yaml

123 lines
2.5 KiB
YAML
Raw Permalink Normal View History

apiVersion: helm.toolkit.fluxcd.io/v2
kind: HelmRelease
metadata:
name: gpu-feature-discovery
namespace: nvidia
spec:
interval: 15m
releaseName: gpu-feature-discovery
targetNamespace: nvidia
chart:
spec:
chart: nvidia-device-plugin
version: 0.20.1
sourceRef:
kind: HelmRepository
name: nvidia-device-plugin
namespace: flux-system
install:
remediation:
retries: 3
upgrade:
remediation:
retries: 3
values:
runtimeClassName: nvidia
resources:
requests:
cpu: 5m
memory: 64Mi
limits:
memory: 128Mi
devicePlugin:
enabled: false
gfd:
enabled: true
securityContext:
privileged: false
allowPrivilegeEscalation: false
capabilities:
drop:
- ALL
nfd:
enabled: false
---
apiVersion: helm.toolkit.fluxcd.io/v2
kind: HelmRelease
metadata:
name: nvidia-device-plugin
namespace: nvidia
spec:
interval: 15m
releaseName: nvidia-device-plugin
targetNamespace: nvidia
dependsOn:
- name: gpu-feature-discovery
chart:
spec:
chart: nvidia-device-plugin
version: 0.20.1
sourceRef:
kind: HelmRepository
name: nvidia-device-plugin
namespace: flux-system
install:
remediation:
retries: 3
upgrade:
remediation:
retries: 3
postRenderers:
- kustomize:
patches:
- target:
kind: DaemonSet
name: nvidia-device-plugin-mps-control-daemon
patch: |
$patch: delete
apiVersion: apps/v1
kind: DaemonSet
metadata:
name: nvidia-device-plugin-mps-control-daemon
values:
runtimeClassName: nvidia
resources:
requests:
cpu: 5m
memory: 64Mi
limits:
memory: 128Mi
nfd:
enabled: false
---
apiVersion: helm.toolkit.fluxcd.io/v2
kind: HelmRelease
metadata:
name: dcgm-exporter
namespace: nvidia
spec:
interval: 15m
releaseName: dcgm-exporter
targetNamespace: nvidia
dependsOn:
- name: nvidia-device-plugin
chart:
spec:
chart: dcgm-exporter
version: 4.8.4
sourceRef:
kind: HelmRepository
name: nvidia-dcgm-exporter
namespace: flux-system
install:
remediation:
retries: 3
upgrade:
remediation:
retries: 3
values:
runtimeClassName: nvidia
nodeSelector:
feature.node.kubernetes.io/pci-10de.present: "true"
serviceMonitor:
enabled: false