feat: stabilize tailscale observability exposure with declarative proxy class

ansible/roles/observability/defaults/main.yml

@@ -20,3 +20,8 @@ loki_enabled: true
 tailscale_oauth_client_id: ""
 tailscale_oauth_client_secret: ""
 tailscale_tailnet: ""
+
+observability_tailscale_expose: true
+grafana_tailscale_hostname: "grafana"
+prometheus_tailscale_hostname: "prometheus"
+tailscale_proxyclass_name: "infra-stable"

ansible/roles/observability/tasks/main.yml

@@ -156,64 +156,66 @@
   changed_when: true
   when: loki_enabled
 
-- name: Configure Grafana for Tailscale access
-  block:
-    - name: Patch Grafana service for Tailscale
-      command: >-
-        kubectl -n {{ observability_namespace }} patch svc kube-prometheus-stack-grafana
-        -p '{"metadata":{"annotations":{"tailscale.com/hostname":"grafana"}},"spec":{"type":"LoadBalancer","loadBalancerClass":"tailscale"}}'
-      register: grafana_patch
-      changed_when: true
-
-    - name: Patch Prometheus service for Tailscale
-      command: >-
-        kubectl -n {{ observability_namespace }} patch svc kube-prometheus-stack-prometheus
-        -p '{"metadata":{"annotations":{"tailscale.com/hostname":"prometheus"}},"spec":{"type":"LoadBalancer","loadBalancerClass":"tailscale"}}'
-      register: prometheus_patch
-      changed_when: true
-
-    - name: Check Tailscale endpoint (IP/hostname) for Grafana
-      shell: >-
-        kubectl -n {{ observability_namespace }} get svc kube-prometheus-stack-grafana
-        -o go-template='{{"{{"}}range .status.loadBalancer.ingress{{"}}"}}{{"{{"}}if .ip{{"}}"}}{{"{{"}}.ip{{"}}"}}{{"{{"}}else{{"}}"}}{{"{{"}}.hostname{{"}}"}}{{"{{"}}end{{"}}"}}{{"{{"}}end{{"}}"}}'
-      register: grafana_lb_ip
-      changed_when: false
-      failed_when: false
-
-    - name: Check Tailscale endpoint (IP/hostname) for Prometheus
-      shell: >-
-        kubectl -n {{ observability_namespace }} get svc kube-prometheus-stack-prometheus
-        -o go-template='{{"{{"}}range .status.loadBalancer.ingress{{"}}"}}{{"{{"}}if .ip{{"}}"}}{{"{{"}}.ip{{"}}"}}{{"{{"}}else{{"}}"}}{{"{{"}}.hostname{{"}}"}}{{"{{"}}end{{"}}"}}{{"{{"}}end{{"}}"}}'
-      register: prometheus_lb_ip
-      changed_when: false
-      failed_when: false
-
-    - name: Warn if Tailscale endpoint assignment is still pending
-      debug:
-        msg: |
-          Tailscale service endpoint assignment is still pending.
-          Grafana endpoint: {{ grafana_lb_ip.stdout | default('') }}
-          Prometheus endpoint: {{ prometheus_lb_ip.stdout | default('') }}
-          Deployment continues; services may become reachable shortly.
-      when: (grafana_lb_ip.stdout | default('') | length == 0) or (prometheus_lb_ip.stdout | default('') | length == 0)
-
-    - name: Show Tailscale access details
-      debug:
-        msg: |
-          Observability stack deployed with Tailscale access!
-
-          Grafana: http://grafana{% if grafana_lb_ip.stdout | default('') | length > 0 %} (or http://{{ grafana_lb_ip.stdout }}){% endif %}
-          Prometheus: http://prometheus{% if prometheus_lb_ip.stdout | default('') | length > 0 %} (or http://{{ prometheus_lb_ip.stdout }}){% endif %}
-
-          Login: admin / {{ grafana_password_effective }}
-
-          Access via:
-          - MagicDNS: http://grafana or http://prometheus (if enabled)
-          - Direct endpoint: {% if grafana_lb_ip.stdout | default('') | length > 0 %}http://{{ grafana_lb_ip.stdout }}{% else %}(pending){% endif %} / {% if prometheus_lb_ip.stdout | default('') | length > 0 %}http://{{ prometheus_lb_ip.stdout }}{% else %}(pending){% endif %}
-          - Tailnet FQDN: http://grafana.{{ tailscale_tailnet | default('tailnet.ts.net') }}
-
-          Note: Ensure Tailscale Kubernetes Operator is installed first
+- name: Check Tailscale service readiness for Grafana
+  command: kubectl -n {{ observability_namespace }} get svc kube-prometheus-stack-grafana -o jsonpath='{.status.conditions[?(@.type=="TailscaleProxyReady")].status}'
+  register: grafana_tailscale_ready
+  changed_when: false
+  failed_when: false
   when:
+    - observability_tailscale_expose | bool
+    - tailscale_operator_ready | default(false) | bool
+
+- name: Check Tailscale service readiness for Prometheus
+  command: kubectl -n {{ observability_namespace }} get svc kube-prometheus-stack-prometheus -o jsonpath='{.status.conditions[?(@.type=="TailscaleProxyReady")].status}'
+  register: prometheus_tailscale_ready
+  changed_when: false
+  failed_when: false
+  when:
+    - observability_tailscale_expose | bool
+    - tailscale_operator_ready | default(false) | bool
+
+- name: Check Tailscale endpoint (IP/hostname) for Grafana
+  shell: >-
+    kubectl -n {{ observability_namespace }} get svc kube-prometheus-stack-grafana
+    -o go-template='{{"{{"}}range .status.loadBalancer.ingress{{"}}"}}{{"{{"}}if .ip{{"}}"}}{{"{{"}}.ip{{"}}"}}{{"{{"}}else{{"}}"}}{{"{{"}}.hostname{{"}}"}}{{"{{"}}end{{"}}"}}{{"{{"}}end{{"}}"}}'
+  register: grafana_lb_ip
+  changed_when: false
+  failed_when: false
+  when:
+    - observability_tailscale_expose | bool
+    - tailscale_operator_ready | default(false) | bool
+
+- name: Check Tailscale endpoint (IP/hostname) for Prometheus
+  shell: >-
+    kubectl -n {{ observability_namespace }} get svc kube-prometheus-stack-prometheus
+    -o go-template='{{"{{"}}range .status.loadBalancer.ingress{{"}}"}}{{"{{"}}if .ip{{"}}"}}{{"{{"}}.ip{{"}}"}}{{"{{"}}else{{"}}"}}{{"{{"}}.hostname{{"}}"}}{{"{{"}}end{{"}}"}}{{"{{"}}end{{"}}"}}'
+  register: prometheus_lb_ip
+  changed_when: false
+  failed_when: false
+  when:
+    - observability_tailscale_expose | bool
+    - tailscale_operator_ready | default(false) | bool
+
+- name: Show Tailscale access details
+  debug:
+    msg: |
+      Observability stack deployed with Tailscale access!
+
+      Grafana: http://{{ grafana_tailscale_hostname }}{% if grafana_lb_ip.stdout | default('') | length > 0 %} (or http://{{ grafana_lb_ip.stdout }}){% endif %}
+      Prometheus: http://{{ prometheus_tailscale_hostname }}{% if prometheus_lb_ip.stdout | default('') | length > 0 %} (or http://{{ prometheus_lb_ip.stdout }}){% endif %}
+
+      Login: admin / {{ grafana_password_effective }}
+
+      Tailscale readiness:
+      - Grafana proxy ready: {{ grafana_tailscale_ready.stdout | default('pending') }}
+      - Prometheus proxy ready: {{ prometheus_tailscale_ready.stdout | default('pending') }}
+
+      Access via:
+      - MagicDNS: http://{{ grafana_tailscale_hostname }} and http://{{ prometheus_tailscale_hostname }}
+      - Tailnet FQDN: http://{{ grafana_tailscale_hostname }}.{{ tailscale_tailnet | default('tailnet.ts.net') }}
+      - Direct endpoint: {% if grafana_lb_ip.stdout | default('') | length > 0 %}http://{{ grafana_lb_ip.stdout }}{% else %}(pending){% endif %} / {% if prometheus_lb_ip.stdout | default('') | length > 0 %}http://{{ prometheus_lb_ip.stdout }}{% else %}(pending){% endif %}
+  when:
+    - observability_tailscale_expose | bool
     - tailscale_operator_ready | default(false) | bool
 
 - name: Show observability access details (fallback)
@@ -230,4 +232,4 @@
       Loki: Disabled
       {% endif %}
   when:
-    - not (tailscale_operator_ready | default(false) | bool)
+    - not (observability_tailscale_expose | bool and (tailscale_operator_ready | default(false) | bool))

ansible/roles/observability/templates/kube-prometheus-stack-values.yaml.j2

@@ -6,8 +6,26 @@ grafana:
     storageClassName: {{ grafana_storage_class }}
     size: {{ grafana_storage_size }}
   service:
+{% if observability_tailscale_expose and (tailscale_operator_ready | default(false)) %}
+    type: LoadBalancer
+    loadBalancerClass: tailscale
+    annotations:
+      tailscale.com/hostname: {{ grafana_tailscale_hostname }}
+      tailscale.com/proxy-class: {{ tailscale_proxyclass_name }}
+{% else %}
     type: ClusterIP
+{% endif %}
 prometheus:
+  service:
+{% if observability_tailscale_expose and (tailscale_operator_ready | default(false)) %}
+    type: LoadBalancer
+    loadBalancerClass: tailscale
+    annotations:
+      tailscale.com/hostname: {{ prometheus_tailscale_hostname }}
+      tailscale.com/proxy-class: {{ tailscale_proxyclass_name }}
+{% else %}
+    type: ClusterIP
+{% endif %}
   prometheusSpec:
     retention: 7d
     storageSpec: