From 96750d5e83302909cbce25a1f35372f5fcb77064 Mon Sep 17 00:00:00 2001
From: Dylan <dystewar@redhat.com>
Date: Wed, 11 Dec 2024 09:41:06 -0500
Subject: [PATCH] nerc-ocp-prod: Add RHOAI GPU accelereatorProfiles

Adds unique acceleratorProfiles for each GPU type in production cluster.
These profiles will show up under accelerators in the RHOAI workbench wizard.

Part of: https://github.com/nerc-project/operations/issues/849
Closes: https://github.com/nerc-project/operations/issues/847
---
 .../a100-acceleratorprofile.yaml                   | 14 ++++++++++++++
 .../rhoai/acceleratorprofiles/kustomization.yaml   |  6 ++++++
 .../v100-acceleratorprofile.yaml                   | 14 ++++++++++++++
 .../nerc-ocp-prod/rhoai/kustomization.yaml         |  1 +
 4 files changed, 35 insertions(+)
 create mode 100644 cluster-scope/overlays/nerc-ocp-prod/rhoai/acceleratorprofiles/a100-acceleratorprofile.yaml
 create mode 100644 cluster-scope/overlays/nerc-ocp-prod/rhoai/acceleratorprofiles/kustomization.yaml
 create mode 100644 cluster-scope/overlays/nerc-ocp-prod/rhoai/acceleratorprofiles/v100-acceleratorprofile.yaml

diff --git a/cluster-scope/overlays/nerc-ocp-prod/rhoai/acceleratorprofiles/a100-acceleratorprofile.yaml b/cluster-scope/overlays/nerc-ocp-prod/rhoai/acceleratorprofiles/a100-acceleratorprofile.yaml
new file mode 100644
index 00000000..4faddd40
--- /dev/null
+++ b/cluster-scope/overlays/nerc-ocp-prod/rhoai/acceleratorprofiles/a100-acceleratorprofile.yaml
@@ -0,0 +1,14 @@
+apiVersion: dashboard.opendatahub.io/v1
+kind: AcceleratorProfile
+metadata:
+  name: nvidia-a100-gpu
+  namespace: redhat-ods-applications
+spec:
+  displayName: NVIDIA A100 GPU
+  enabled: true
+  identifier: nvidia.com/gpu
+  tolerations:
+    - effect: NoSchedule
+      key: nvidia.com/gpu.product
+      operator: Equal
+      value: NVIDIA-A100-SXM4-40GB
diff --git a/cluster-scope/overlays/nerc-ocp-prod/rhoai/acceleratorprofiles/kustomization.yaml b/cluster-scope/overlays/nerc-ocp-prod/rhoai/acceleratorprofiles/kustomization.yaml
new file mode 100644
index 00000000..8094ae5a
--- /dev/null
+++ b/cluster-scope/overlays/nerc-ocp-prod/rhoai/acceleratorprofiles/kustomization.yaml
@@ -0,0 +1,6 @@
+apiVersion: kustomize.config.k8s.io/v1beta1
+kind: Kustomization
+
+resources:
+  - a100-acceleratorprofile.yaml
+  - v100-acceleratorprofile.yaml
diff --git a/cluster-scope/overlays/nerc-ocp-prod/rhoai/acceleratorprofiles/v100-acceleratorprofile.yaml b/cluster-scope/overlays/nerc-ocp-prod/rhoai/acceleratorprofiles/v100-acceleratorprofile.yaml
new file mode 100644
index 00000000..f8e9484a
--- /dev/null
+++ b/cluster-scope/overlays/nerc-ocp-prod/rhoai/acceleratorprofiles/v100-acceleratorprofile.yaml
@@ -0,0 +1,14 @@
+apiVersion: dashboard.opendatahub.io/v1
+kind: AcceleratorProfile
+metadata:
+  name: nvidia-v100-gpu
+  namespace: redhat-ods-applications
+spec:
+  displayName: NVIDIA V100 GPU
+  enabled: true
+  identifier: nvidia.com/gpu
+  tolerations:
+    - effect: NoSchedule
+      key: nvidia.com/gpu.product
+      operator: Equal
+      value: Tesla-V100-PCIE-32GB
diff --git a/cluster-scope/overlays/nerc-ocp-prod/rhoai/kustomization.yaml b/cluster-scope/overlays/nerc-ocp-prod/rhoai/kustomization.yaml
index 3350eb1f..439e040b 100644
--- a/cluster-scope/overlays/nerc-ocp-prod/rhoai/kustomization.yaml
+++ b/cluster-scope/overlays/nerc-ocp-prod/rhoai/kustomization.yaml
@@ -4,3 +4,4 @@ resources:
 - odhdashboardconfigs/odh-dashboard-config.yaml
 - imagestreams/ucsls-f24-imagestream.yaml
 - imagestreams/oauth-proxy.yaml
+- acceleratorprofiles