Auto-update README with current k8s applications

Generated by CI/CD workflow on 2026-03-07 21:24:30

This PR updates the README.md file with the current list of applications found in the k8s/ directory structure.

k8s/apps/comfyui/app.yaml

@@ -4,7 +4,7 @@ metadata:
   name: comfyui
   namespace: argocd
 spec:
-  project: apps
+  project: core
   destination:
     namespace: comfyui
     server: https://kubernetes.default.svc

k8s/apps/comfyui/deployment.yaml

@@ -16,30 +16,18 @@ spec:
         app: comfyui
     spec:
       runtimeClassName: nvidia
-      tolerations:
-        - key: workload
-          operator: Equal
-          value: desktop
-          effect: NoSchedule
       nodeSelector:
         kubernetes.io/hostname: uk-desktop.tail2fe2d.ts.net
       # Fix permissions mismatch usually happening when mapping host paths
       securityContext:
         runAsUser: 0
-      initContainers:
-        - name: create-data-dir
-          image: busybox
-          command: ["sh", "-c", "mkdir -p /host.data && chown -R 1000:1000 /host.data"]
-          volumeMounts:
-            - name: data
-              mountPath: /host.data
       containers:
         - name: comfyui
-          image: runpod/comfyui:latest-5090
+          image: ghcr.io/ai-dock/comfyui:latest-cuda
           imagePullPolicy: IfNotPresent
           env:
-            - name: COMFYUI_PORT
-              value: "8188"
+            - name: CLI_ARGS
+              value: "--listen 0.0.0.0 --port 8188"
           ports:
             - containerPort: 8188
               name: http

k8s/core/gpu/values.yaml

@@ -10,14 +10,3 @@ tolerations:
 runtimeClassName: nvidia
 
 setAsDefault: false
-
-config:
-  default: any
-  map:
-    any: |-
-      version: v1
-      sharing:
-        timeSlicing:
-          resources:
-          - name: nvidia.com/gpu
-            replicas: 4

k8s/core/prom-stack/grafana-alerting-configmap.yaml

@@ -260,7 +260,7 @@ data:
                   to: 0
                 datasourceUid: P76F38748CEC837F0
                 model:
-                  expr: 'node_load15 / on(instance) group_left count by(instance)(node_cpu_seconds_total{mode="idle"})'
+                  expr: 'node_load5 / on(instance) group_left count by(instance)(node_cpu_seconds_total{mode="idle"})'
                   refId: A
                   intervalMs: 1000
                   maxDataPoints: 43200
@@ -273,7 +273,7 @@ data:
                   conditions:
                     - evaluator:
                         params:
-                          - 2
+                          - 0.8
                         type: gt
                       operator:
                         type: and
@@ -283,16 +283,16 @@ data:
                     type: __expr__
                     uid: __expr__
                   expression: A
-                  reducer: last
+                  reducer: max
                   refId: B
                   type: reduce
             noDataState: NoData
             execErrState: Alerting
-            for: 15m
+            for: 5m
             annotations:
               node: '{{ $labels.instance }}'
               load_average: '{{ printf "%.2f" $values.A }}'
-              summary: 'Node load average is critically high relative to CPU count'
+              summary: 'Node load average is high relative to CPU count'
             labels:
               severity: warning
 

k8s/core/prom-stack/grafana-values.yaml

@@ -0,0 +1,85 @@
+envFromSecret: grafana-admin
+nodeSelector:
+  kubernetes.io/hostname: master.tail2fe2d.ts.net
+
+admin:
+  existingSecret: grafana-admin
+  userKey: username
+  passwordKey: password
+
+grafana.ini:
+  auth:
+    signout_redirect_url: https://idm.hexor.cy/application/o/grafana/end-session/
+    # oauth_auto_login: true
+  auth.generic_oauth:
+    name: authentik
+    enabled: true
+    scopes: "openid profile email"
+    auth_url: https://idm.hexor.cy/application/o/authorize/
+    token_url: https://idm.hexor.cy/application/o/token/
+    api_url: https://idm.hexor.cy/application/o/userinfo/
+    role_attribute_path: >-
+      contains(groups, 'Grafana Admin') && 'Admin' ||
+      contains(groups, 'Grafana Editors') && 'Editor' ||
+      contains(groups, 'Grafana Viewer') && 'Viewer'
+  database:
+    type: postgres
+    host: psql.psql.svc:5432
+    name: grafana
+    user: grafana
+    ssl_mode: disable
+
+datasources:
+  datasources.yaml:
+    apiVersion: 1
+    datasources:
+      - name: Prometheus Local
+        type: prometheus
+        url: http://prometheus-kube-prometheus-prometheus.prometheus.svc:9090
+        access: proxy
+        isDefault: true
+      - name: Loki
+        type: loki
+        url: http://loki-gateway.prometheus.svc:80
+        access: proxy
+
+ingress:
+  enabled: true
+  ingressClassName: traefik
+  annotations:
+    cert-manager.io/cluster-issuer: letsencrypt
+    traefik.ingress.kubernetes.io/router.middlewares: kube-system-https-redirect@kubernetescrd
+  hosts:
+    - gf.hexor.cy
+  tls:
+    - secretName: grafana-tls
+      hosts:
+        - '*.hexor.cy'
+
+extraConfigmapMounts:
+  - name: grafana-alerting-rules
+    mountPath: /etc/grafana/provisioning/alerting/rules.yaml
+    configMap: grafana-alerting
+    subPath: rules.yaml
+    readOnly: true
+  - name: grafana-alerting-contactpoints
+    mountPath: /etc/grafana/provisioning/alerting/contactpoints.yaml
+    configMap: grafana-alerting
+    subPath: contactpoints.yaml
+    readOnly: true
+  - name: grafana-alerting-policies
+    mountPath: /etc/grafana/provisioning/alerting/policies.yaml
+    configMap: grafana-alerting
+    subPath: policies.yaml
+    readOnly: true
+
+envValueFrom:
+  TELEGRAM_BOT_TOKEN:
+    secretKeyRef:
+      name: grafana-telegram
+      key: bot-token
+  TELEGRAM_CHAT_ID:
+    secretKeyRef:
+      name: grafana-telegram
+      key: chat-id
+

k8s/core/prom-stack/kustomization.yaml

@@ -16,6 +16,14 @@ helmCharts:
     valuesFile: prom-values.yaml
     includeCRDs: true
 
+  - name: grafana
+    repo: https://grafana.github.io/helm-charts
+    version: 10.2.0
+    releaseName: grafana
+    namespace: prometheus
+    valuesFile: grafana-values.yaml
+    includeCRDs: true
+
   - name: loki
     repo: https://grafana.github.io/helm-charts
     version: 6.29.0

k8s/core/prom-stack/prom-values.yaml

@@ -1,4 +1,5 @@
-
+grafana:
+  enabled: false
 
 alertmanager:
   config:
@@ -91,88 +92,3 @@ prometheus:
             requests:
               storage: 400Gi
 
-grafana:
-  enabled: true
-  
-  serviceAccount:
-    create: true
-    name: "prom-grafana-sa"
-
-  envFromSecret: grafana-admin
-  nodeSelector:
-    kubernetes.io/hostname: master.tail2fe2d.ts.net
-
-  admin:
-    existingSecret: grafana-admin
-    userKey: username
-    passwordKey: password
-
-  grafana.ini:
-    auth:
-      signout_redirect_url: https://idm.hexor.cy/application/o/grafana/end-session/
-    auth.generic_oauth:
-      name: authentik
-      enabled: true
-      scopes: "openid profile email"
-      auth_url: https://idm.hexor.cy/application/o/authorize/
-      token_url: https://idm.hexor.cy/application/o/token/
-      api_url: https://idm.hexor.cy/application/o/userinfo/
-      role_attribute_path: >-
-        contains(groups, 'Grafana Admin') && 'Admin' ||
-        contains(groups, 'Grafana Editors') && 'Editor' ||
-        contains(groups, 'Grafana Viewer') && 'Viewer'
-    database:
-      type: postgres
-      host: psql.psql.svc:5432
-      name: grafana
-      user: grafana
-      ssl_mode: disable
-
-  # The Loki datasource config needs to be preserved, 
-  # but instead of "datasources.datasources.yaml", we define it like this for the prometheus-stack chart:
-  additionalDataSources:
-    - name: Loki
-      type: loki
-      url: http://loki-gateway.prometheus.svc:80
-      access: proxy
-      orgId: 1
-
-  ingress:
-    enabled: true
-    ingressClassName: traefik
-    annotations:
-      cert-manager.io/cluster-issuer: letsencrypt
-      traefik.ingress.kubernetes.io/router.middlewares: kube-system-https-redirect@kubernetescrd
-    hosts:
-      - gf.hexor.cy
-    tls:
-      - secretName: grafana-tls
-        hosts:
-          - '*.hexor.cy'
-
-  extraConfigmapMounts:
-    - name: grafana-alerting-rules
-      mountPath: /etc/grafana/provisioning/alerting/rules.yaml
-      configMap: grafana-alerting
-      subPath: rules.yaml
-      readOnly: true
-    - name: grafana-alerting-contactpoints
-      mountPath: /etc/grafana/provisioning/alerting/contactpoints.yaml
-      configMap: grafana-alerting
-      subPath: contactpoints.yaml
-      readOnly: true
-    - name: grafana-alerting-policies
-      mountPath: /etc/grafana/provisioning/alerting/policies.yaml
-      configMap: grafana-alerting
-      subPath: policies.yaml
-      readOnly: true
-
-  envValueFrom:
-    TELEGRAM_BOT_TOKEN:
-      secretKeyRef:
-        name: grafana-telegram
-        key: bot-token
-    TELEGRAM_CHAT_ID:
-      secretKeyRef:
-        name: grafana-telegram
-        key: chat-id