Add roles for GPU operator and monitoring charts

playbooks/roles/charts/gpu-operator/files/setup.sh

@@ -0,0 +1,11 @@
+#!/bin/bash
+helm upgrade --install gpu-operator nvidia/gpu-operator \
+  --namespace gpu-operator \
+  --create-namespace \
+  --set nodeSelector.kubernetes.io/gpu="true" \
+  --set driver.enabled=true \
+  --set toolkit.enabled=true \
+  --set devicePlugin.enabled=true \
+  --set operator.runtimeClass="nvidia-container-runtime" \
+  --set operator.defaultRuntime=containerd \
+  --set containerRuntime.socketPath=/var/snap/microk8s/common/run/containerd.sock

playbooks/roles/charts/gpu-operator/tasks/main.yml

@@ -0,0 +1,3 @@
+- name: Install GPU Operator
+  script: files/setup.sh
+  when: is_primary | bool

playbooks/roles/charts/helm-repos/tasks/main.yml

@@ -0,0 +1,9 @@
+- name: Enable community plugins and third-party helm charts
+  shell: |
+    helm repo add kubernetes-dashboard https://kubernetes.github.io/dashboard/ || true
+    helm repo add nvidia https://helm.ngc.nvidia.com/nvidia || true
+    helm repo add prometheus-community https://prometheus-community.github.io/helm-charts || true
+    helm repo add metrics-server https://kubernetes-sigs.github.io/metrics-server/ || true
+    helm repo update
+  ignore_errors: yes
+  when: is_primary | bool

playbooks/roles/charts/kubernetes-dashboard/files/setup.sh

@@ -0,0 +1,11 @@
+#!/bin/bash
+helm upgrade --install kubernetes-dashboard kubernetes-dashboard/kubernetes-dashboard \
+  --create-namespace \
+  --namespace kubernetes-dashboard \
+  --set app.scheduling.nodeSelector."kubernetes\.io/hostname"=$1 \
+  --set auth.nodeSelector."kubernetes\.io/hostname"=$1 \
+  --set api.nodeSelector."kubernetes\.io/hostname"=$1 \
+  --set web.nodeSelector."kubernetes\.io/hostname"=$1 \
+  --set metricsScraper.nodeSelector."kubernetes\.io/hostname"=$1 \
+  --set kong.nodeSelector."kubernetes\.io/hostname"=$1 \
+  --set persistence.enabled=false

playbooks/roles/charts/kubernetes-dashboard/tasks/main.yml

@@ -0,0 +1,3 @@
+- name: Install kubernetes dashboard
+  script: files/setup.sh {{ inventory_hostname }}
+  when: is_primary | bool

playbooks/roles/charts/metrics-server/files/setup.sh

@@ -0,0 +1,5 @@
+#!/bin/bash
+helm upgrade --install metrics-server metrics-server/metrics-server \
+  --namespace kube-system \
+  --set nodeSelector."kubernetes\.io/hostname"=$1 \
+  --set persistence.enabled=false

playbooks/roles/charts/metrics-server/tasks/main.yml

@@ -0,0 +1,3 @@
+- name: Install metrics server
+  script: files/setup.sh {{ inventory_hostname }}
+  when: is_primary | bool

playbooks/roles/charts/prometheus/files/setup.sh

@@ -0,0 +1,13 @@
+#!/bin/bash
+helm upgrade --install prometheus prometheus-community/prometheus \
+  --namespace chutes \
+  --create-namespace \
+  --set server.persistentVolume.enabled=false \
+  --set alertmanager.persistentVolume.enabled=false \
+  --set prometheus-pushgateway.persistentVolume.enabled=false \
+  --set prometheus-server.persistentVolume.enabled=false \
+  --set alertmanager.persistence.enabled=false \
+  --set server.nodeSelector."kubernetes\.io/hostname"=$1 \
+  --set alertmanager.nodeSelector."kubernetes\.io/hostname"=$1 \
+  --set pushgateway.nodeSelector."kubernetes\.io/hostname"=$1 \
+  --set kubeStateMetrics.nodeSelector."kubernetes\.io/hostname"=$1

playbooks/roles/charts/prometheus/tasks/main.yml

@@ -0,0 +1,3 @@
+- name: Install Prometheus
+  script: files/setup.sh {{ inventory_hostname }}
+  when: is_primary | bool