f7c3a338a5
Phase 2 of platform stack split. 5 more modules extracted into independent stacks. All applied successfully with zero destroys. Cloudflared now reads k8s_users from Vault directly to compute user_domains. Woodpecker pipeline runs all 8 extracted stacks in parallel. Memory bumped to 6Gi for 9 concurrent TF processes. Platform reduced from 27 to 19 modules.
43 lines
1.1 KiB
YAML
43 lines
1.1 KiB
YAML
driver:
|
|
enabled: true
|
|
# repository: nvcr.io/nvidia/driver
|
|
# choose a driver version compatible with your GPU + CUDA 12.x (example)
|
|
# NVIDIA GPU driver - https://docs.nvidia.com/datacenter/cloud-native/gpu-operator/latest/platform-support.html#known-issue
|
|
# https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/
|
|
# 13.x >= 580
|
|
# 12.x >= 525, <580
|
|
# 11.x >= 450, <525
|
|
#
|
|
# Delete the cluster policy before each change
|
|
# version: "575.57.08" # CUDA 12.9
|
|
version: "570.195.03" # CUDA 12.8
|
|
upgradePolicy:
|
|
autoUpgrade: false
|
|
|
|
devicePlugin:
|
|
config:
|
|
name: time-slicing-config
|
|
|
|
# DCGM Exporter - reduced from 2560Mi to 1536Mi based on VPA upper bound of 1459Mi (1.05x margin)
|
|
dcgmExporter:
|
|
resources:
|
|
requests:
|
|
memory: "1536Mi"
|
|
limits:
|
|
memory: "1536Mi"
|
|
|
|
# CUDA Validator - reduced from 1024Mi to 256Mi (one-shot job)
|
|
validator:
|
|
resources:
|
|
requests:
|
|
memory: "256Mi"
|
|
limits:
|
|
memory: "256Mi"
|
|
|
|
# Tolerate GPU node taint for all GPU operator components
|
|
daemonsets:
|
|
tolerations:
|
|
- key: "nvidia.com/gpu"
|
|
operator: "Equal"
|
|
value: "true"
|
|
effect: "NoSchedule"
|