From a7bc2a425e1773bc33c1b08e46b25207a996d762 Mon Sep 17 00:00:00 2001 From: Brad Stein Date: Mon, 25 Aug 2025 23:41:36 -0500 Subject: [PATCH] daemonset restructuring --- infrastructure/core/base/kustomization.yaml | 6 ++++++ .../{ => base}/priorityclass/kustomization.yaml | 2 +- .../core/{ => base}/priorityclass/scavenger.yaml | 2 +- .../core/{ => base}/storageclass/asteria.yaml | 2 +- .../core/{ => base}/storageclass/astreae.yaml | 2 +- .../{ => base}/storageclass/kustomization.yaml | 2 +- .../device-plugin-jetson/kustomization.yaml | 4 ---- .../device-plugin-minipc/kustomization.yaml | 4 ---- .../kustomization-device-plugin-jetson.yaml | 15 --------------- .../kustomization-device-plugin-minipc.yaml | 15 --------------- .../device-plugin-jetson/daemonset.yaml | 2 +- .../gpu/device-plugin-jetson/kustomization.yaml | 5 +++++ .../device-plugin-minipc/daemonset.yaml | 2 +- .../gpu/device-plugin-minipc/kustomization.yaml | 5 +++++ .../gpu/profiles/jetson-only/kustomization.yaml | 5 +++++ .../profiles/minipc-and-jetson/kustomization.yaml | 6 ++++++ .../gpu/profiles/minipc-only/kustomization.yaml | 5 +++++ infrastructure/core/kustomization.yaml | 10 ++++------ .../flux-system/kustomization-core.yaml | 10 ++++++++-- 19 files changed, 51 insertions(+), 53 deletions(-) create mode 100644 infrastructure/core/base/kustomization.yaml rename infrastructure/core/{ => base}/priorityclass/kustomization.yaml (61%) rename infrastructure/core/{ => base}/priorityclass/scavenger.yaml (77%) rename infrastructure/core/{ => base}/storageclass/asteria.yaml (84%) rename infrastructure/core/{ => base}/storageclass/astreae.yaml (87%) rename infrastructure/core/{ => base}/storageclass/kustomization.yaml (64%) delete mode 100644 infrastructure/core/daemonsets/device-plugin-jetson/kustomization.yaml delete mode 100644 infrastructure/core/daemonsets/device-plugin-minipc/kustomization.yaml delete mode 100644 infrastructure/core/daemonsets/kustomization-device-plugin-jetson.yaml delete mode 100644 infrastructure/core/daemonsets/kustomization-device-plugin-minipc.yaml rename infrastructure/core/{daemonsets => gpu}/device-plugin-jetson/daemonset.yaml (94%) create mode 100644 infrastructure/core/gpu/device-plugin-jetson/kustomization.yaml rename infrastructure/core/{daemonsets => gpu}/device-plugin-minipc/daemonset.yaml (95%) create mode 100644 infrastructure/core/gpu/device-plugin-minipc/kustomization.yaml create mode 100644 infrastructure/core/gpu/profiles/jetson-only/kustomization.yaml create mode 100644 infrastructure/core/gpu/profiles/minipc-and-jetson/kustomization.yaml create mode 100644 infrastructure/core/gpu/profiles/minipc-only/kustomization.yaml diff --git a/infrastructure/core/base/kustomization.yaml b/infrastructure/core/base/kustomization.yaml new file mode 100644 index 0000000..07817fd --- /dev/null +++ b/infrastructure/core/base/kustomization.yaml @@ -0,0 +1,6 @@ +# infrastructure/core/base/kustomization.yaml +apiVersion: kustomize.config.k8s.io/v1beta1 +kind: Kustomization +resources: + - storageclass + - priorityclass diff --git a/infrastructure/core/priorityclass/kustomization.yaml b/infrastructure/core/base/priorityclass/kustomization.yaml similarity index 61% rename from infrastructure/core/priorityclass/kustomization.yaml rename to infrastructure/core/base/priorityclass/kustomization.yaml index 19b2a7a..9f2960d 100644 --- a/infrastructure/core/priorityclass/kustomization.yaml +++ b/infrastructure/core/base/priorityclass/kustomization.yaml @@ -1,4 +1,4 @@ -# infrastructure/core/priorityclass/kustomization.yaml +# infrastructure/core/base/priorityclass/kustomization.yaml apiVersion: kustomize.config.k8s.io/v1beta1 kind: Kustomization resources: diff --git a/infrastructure/core/priorityclass/scavenger.yaml b/infrastructure/core/base/priorityclass/scavenger.yaml similarity index 77% rename from infrastructure/core/priorityclass/scavenger.yaml rename to infrastructure/core/base/priorityclass/scavenger.yaml index 1627535..ac09ba4 100644 --- a/infrastructure/core/priorityclass/scavenger.yaml +++ b/infrastructure/core/base/priorityclass/scavenger.yaml @@ -1,4 +1,4 @@ -# infrastructure/core/priorityclass/scavenger.yaml +# infrastructure/core/base/priorityclass/scavenger.yaml apiVersion: scheduling.k8s.io/v1 kind: PriorityClass metadata: diff --git a/infrastructure/core/storageclass/asteria.yaml b/infrastructure/core/base/storageclass/asteria.yaml similarity index 84% rename from infrastructure/core/storageclass/asteria.yaml rename to infrastructure/core/base/storageclass/asteria.yaml index d3c86fe..eb5947f 100644 --- a/infrastructure/core/storageclass/asteria.yaml +++ b/infrastructure/core/base/storageclass/asteria.yaml @@ -1,4 +1,4 @@ -# infrastructure/core/storageclass/asteria.yaml +# infrastructure/core/base/storageclass/asteria.yaml apiVersion: storage.k8s.io/v1 kind: StorageClass metadata: diff --git a/infrastructure/core/storageclass/astreae.yaml b/infrastructure/core/base/storageclass/astreae.yaml similarity index 87% rename from infrastructure/core/storageclass/astreae.yaml rename to infrastructure/core/base/storageclass/astreae.yaml index ec9201a..f169261 100644 --- a/infrastructure/core/storageclass/astreae.yaml +++ b/infrastructure/core/base/storageclass/astreae.yaml @@ -1,4 +1,4 @@ -# infrastructure/core/storageclass/astreae.yaml +# infrastructure/core/base/storageclass/astreae.yaml apiVersion: storage.k8s.io/v1 kind: StorageClass metadata: diff --git a/infrastructure/core/storageclass/kustomization.yaml b/infrastructure/core/base/storageclass/kustomization.yaml similarity index 64% rename from infrastructure/core/storageclass/kustomization.yaml rename to infrastructure/core/base/storageclass/kustomization.yaml index 28dc29d..e45b4bc 100644 --- a/infrastructure/core/storageclass/kustomization.yaml +++ b/infrastructure/core/base/storageclass/kustomization.yaml @@ -1,4 +1,4 @@ -# infrastructure/core/storageclass/kustomization.yaml +# infrastructure/core/base/storageclass/kustomization.yaml apiVersion: kustomize.config.k8s.io/v1beta1 kind: Kustomization resources: diff --git a/infrastructure/core/daemonsets/device-plugin-jetson/kustomization.yaml b/infrastructure/core/daemonsets/device-plugin-jetson/kustomization.yaml deleted file mode 100644 index 5beccfa..0000000 --- a/infrastructure/core/daemonsets/device-plugin-jetson/kustomization.yaml +++ /dev/null @@ -1,4 +0,0 @@ -# infrastructure/core/daemonsets/device-plugin-jetson/kustomization.yaml -apiVersion: kustomize.config.k8s.io/v1beta1 -kind: Kustomization -resources: [daemonset.yaml] diff --git a/infrastructure/core/daemonsets/device-plugin-minipc/kustomization.yaml b/infrastructure/core/daemonsets/device-plugin-minipc/kustomization.yaml deleted file mode 100644 index a0321fc..0000000 --- a/infrastructure/core/daemonsets/device-plugin-minipc/kustomization.yaml +++ /dev/null @@ -1,4 +0,0 @@ -# infrastructure/core/daemonsets/device-plugin-minipc/kustomization.yaml -apiVersion: kustomize.config.k8s.io/v1beta1 -kind: Kustomization -resources: [daemonset.yaml] diff --git a/infrastructure/core/daemonsets/kustomization-device-plugin-jetson.yaml b/infrastructure/core/daemonsets/kustomization-device-plugin-jetson.yaml deleted file mode 100644 index b240edc..0000000 --- a/infrastructure/core/daemonsets/kustomization-device-plugin-jetson.yaml +++ /dev/null @@ -1,15 +0,0 @@ -# infrastructure/core/daemonsets/kustomization-device-plugin-jetson.yaml -apiVersion: kustomize.toolkit.fluxcd.io/v1 -kind: Kustomization -metadata: - name: nvidia-device-plugin-jetson - namespace: flux-system -spec: - suspend: true - interval: 10m - path: ./infrastructure/core/daemonsets/device-plugin-jetson - prune: true - sourceRef: - kind: GitRepository - name: flux-system - namespace: flux-system diff --git a/infrastructure/core/daemonsets/kustomization-device-plugin-minipc.yaml b/infrastructure/core/daemonsets/kustomization-device-plugin-minipc.yaml deleted file mode 100644 index 1864223..0000000 --- a/infrastructure/core/daemonsets/kustomization-device-plugin-minipc.yaml +++ /dev/null @@ -1,15 +0,0 @@ -# infrastructure/core/daemonsets/kustomization-device-plugin-minipc.yaml -apiVersion: kustomize.toolkit.fluxcd.io/v1 -kind: Kustomization -metadata: - name: nvidia-device-plugin-minipc - namespace: flux-system -spec: - suspend: false - interval: 10m - path: ./infrastructure/core/daemonsets/device-plugin-minipc - prune: true - sourceRef: - kind: GitRepository - name: flux-system - namespace: flux-system diff --git a/infrastructure/core/daemonsets/device-plugin-jetson/daemonset.yaml b/infrastructure/core/gpu/device-plugin-jetson/daemonset.yaml similarity index 94% rename from infrastructure/core/daemonsets/device-plugin-jetson/daemonset.yaml rename to infrastructure/core/gpu/device-plugin-jetson/daemonset.yaml index 642ff0e..fbb5b5c 100644 --- a/infrastructure/core/daemonsets/device-plugin-jetson/daemonset.yaml +++ b/infrastructure/core/gpu/device-plugin-jetson/daemonset.yaml @@ -1,4 +1,4 @@ -# infrastructure/core/daemonsets/device-plugin-jetson/daemonset.yaml +# infrastructure/core/gpu/daemonsets/device-plugin-jetson/daemonset.yaml apiVersion: apps/v1 kind: DaemonSet metadata: diff --git a/infrastructure/core/gpu/device-plugin-jetson/kustomization.yaml b/infrastructure/core/gpu/device-plugin-jetson/kustomization.yaml new file mode 100644 index 0000000..10b4796 --- /dev/null +++ b/infrastructure/core/gpu/device-plugin-jetson/kustomization.yaml @@ -0,0 +1,5 @@ +# infrastructure/core/gpu/daemonsets/device-plugin-jetson/kustomization.yaml +apiVersion: kustomize.config.k8s.io/v1beta1 +kind: Kustomization +resources: + - daemonset.yaml diff --git a/infrastructure/core/daemonsets/device-plugin-minipc/daemonset.yaml b/infrastructure/core/gpu/device-plugin-minipc/daemonset.yaml similarity index 95% rename from infrastructure/core/daemonsets/device-plugin-minipc/daemonset.yaml rename to infrastructure/core/gpu/device-plugin-minipc/daemonset.yaml index c221a67..13a1613 100644 --- a/infrastructure/core/daemonsets/device-plugin-minipc/daemonset.yaml +++ b/infrastructure/core/gpu/device-plugin-minipc/daemonset.yaml @@ -1,4 +1,4 @@ -# infrastructure/core/daemonsets/device-plugin-minipc/daemonset.yaml +# infrastructure/core/gpu/daemonsets/device-plugin-minipc/daemonset.yaml apiVersion: apps/v1 kind: DaemonSet metadata: diff --git a/infrastructure/core/gpu/device-plugin-minipc/kustomization.yaml b/infrastructure/core/gpu/device-plugin-minipc/kustomization.yaml new file mode 100644 index 0000000..d05c88d --- /dev/null +++ b/infrastructure/core/gpu/device-plugin-minipc/kustomization.yaml @@ -0,0 +1,5 @@ +# infrastructure/core/gpu/daemonsets/device-plugin-minipc/kustomization.yaml +apiVersion: kustomize.config.k8s.io/v1beta1 +kind: Kustomization +resources: + - daemonset.yaml diff --git a/infrastructure/core/gpu/profiles/jetson-only/kustomization.yaml b/infrastructure/core/gpu/profiles/jetson-only/kustomization.yaml new file mode 100644 index 0000000..1a43f08 --- /dev/null +++ b/infrastructure/core/gpu/profiles/jetson-only/kustomization.yaml @@ -0,0 +1,5 @@ +# infrastructure/core/gpu/daemonsets/profiles/jetson-only/kustomization.yaml +apiVersion: kustomize.config.k8s.io/v1beta1 +kind: Kustomization +resources: + - ../../device-plugin-jetson diff --git a/infrastructure/core/gpu/profiles/minipc-and-jetson/kustomization.yaml b/infrastructure/core/gpu/profiles/minipc-and-jetson/kustomization.yaml new file mode 100644 index 0000000..437a621 --- /dev/null +++ b/infrastructure/core/gpu/profiles/minipc-and-jetson/kustomization.yaml @@ -0,0 +1,6 @@ +# infrastructure/core/gpu/daemonsets/profiles/minipc-and-jetson/kustomization.yaml +apiVersion: kustomize.config.k8s.io/v1beta1 +kind: Kustomization +resources: + - ../../device-plugin-minipc + - ../../device-plugin-jetson diff --git a/infrastructure/core/gpu/profiles/minipc-only/kustomization.yaml b/infrastructure/core/gpu/profiles/minipc-only/kustomization.yaml new file mode 100644 index 0000000..dab4baa --- /dev/null +++ b/infrastructure/core/gpu/profiles/minipc-only/kustomization.yaml @@ -0,0 +1,5 @@ +# infrastructure/core/gpu/daemonsets/profiles/minipc-only/kustomization.yaml +apiVersion: kustomize.config.k8s.io/v1beta1 +kind: Kustomization +resources: + - ../../device-plugin-minipc diff --git a/infrastructure/core/kustomization.yaml b/infrastructure/core/kustomization.yaml index 1fe2dcb..4ab7bbe 100644 --- a/infrastructure/core/kustomization.yaml +++ b/infrastructure/core/kustomization.yaml @@ -2,9 +2,7 @@ apiVersion: kustomize.config.k8s.io/v1beta1 kind: Kustomization resources: - - ./storageclass - - ./priorityclass - - ./daemonsets/device-plugin-minipc - - ./daemonsets/device-plugin-jetson - # - ./daemonsets/kustomization-device-plugin-minipc.yaml - # - ./daemonsets/kustomization-device-plugin-jetson.yaml + - base + - gpu/profiles/jetson-only + # - gpu/profiles/minipc-and-jetson + # - gpu/profiles/minipc-only diff --git a/infrastructure/flux-system/kustomization-core.yaml b/infrastructure/flux-system/kustomization-core.yaml index 38b9409..d3947b5 100644 --- a/infrastructure/flux-system/kustomization-core.yaml +++ b/infrastructure/flux-system/kustomization-core.yaml @@ -7,10 +7,16 @@ metadata: spec: interval: 10m path: ./infrastructure/core - prune: false + prune: true sourceRef: kind: GitRepository name: flux-system namespace: flux-system wait: true - \ No newline at end of file + + # Only wait for the NVIDIA device-plugin DaemonSet on titan-22 + healthChecks: + - apiVersion: apps/v1 + kind: DaemonSet + name: nvidia-device-plugin-minipc + namespace: kube-system