feat: pds updates with telemetry and branding changes

playbook.yml

@@ -7,7 +7,7 @@
   vars:
     pds_data_dir: /pds
     pds_image_tag: "0.4"
-    pds_blob_upload_limit: 104857600
+    pds_blob_upload_limit: 314572800
     pds_rate_limits_enabled: true
     pds_invite_required: true
     pds_backup_period: 24h
@@ -16,6 +16,7 @@
     pds_backup_s3_endpoint: https://s3.bhs.io.cloud.ovh.net
     pds_backup_s3_access_key: "{{ pds_blobstore_s3_access_key }}"
     pds_backup_s3_secret_key: "{{ pds_blobstore_s3_secret_key }}"
+    pds_grafana_image_tag: "13.1.1"
 
   tasks:
     - name: Run PDS installer (first time only)
@@ -135,6 +136,71 @@
       changed_when: false
       failed_when: pds_backup_exit_code.stdout != "0"
 
+    # ── Monitoring stack (telemetry) ──────────────────────────────
+    # Deploys Prometheus + Grafana + node_exporter when pds_telemetry_enabled is true.
+    # All services bind to 127.0.0.1 — access via SSH tunnel:
+    #   ssh -L 3001:localhost:3001 sysadmin@<server>
+    # Then open http://localhost:3001 (admin/admin, change on first login)
+
+    - name: Create monitoring directories
+      file:
+        path: "{{ item }}"
+        state: directory
+        mode: "0755"
+      loop:
+        - "{{ pds_data_dir }}/monitoring"
+        - "{{ pds_data_dir }}/monitoring/grafana/provisioning/datasources"
+        - "{{ pds_data_dir }}/monitoring/grafana/provisioning/dashboards"
+        - "{{ pds_data_dir }}/monitoring/dashboards"
+      when: pds_telemetry_enabled is defined and pds_telemetry_enabled | bool
+
+    - name: Deploy monitoring compose
+      template:
+        src: templates/monitoring-compose.yaml.j2
+        dest: "{{ pds_data_dir }}/monitoring/compose.yaml"
+        mode: "0644"
+      when: pds_telemetry_enabled is defined and pds_telemetry_enabled | bool
+      notify: restart monitoring
+
+    - name: Deploy Prometheus config
+      template:
+        src: templates/prometheus.yml.j2
+        dest: "{{ pds_data_dir }}/monitoring/prometheus.yml"
+        mode: "0644"
+      when: pds_telemetry_enabled is defined and pds_telemetry_enabled | bool
+      notify: restart monitoring
+
+    - name: Deploy Grafana datasource provisioning
+      template:
+        src: templates/grafana-datasources.yml.j2
+        dest: "{{ pds_data_dir }}/monitoring/grafana/provisioning/datasources/prometheus.yml"
+        mode: "0644"
+      when: pds_telemetry_enabled is defined and pds_telemetry_enabled | bool
+      notify: restart monitoring
+
+    - name: Deploy Grafana dashboard provisioning
+      template:
+        src: templates/grafana-dashboards.yml.j2
+        dest: "{{ pds_data_dir }}/monitoring/grafana/provisioning/dashboards/pds.yml"
+        mode: "0644"
+      when: pds_telemetry_enabled is defined and pds_telemetry_enabled | bool
+      notify: restart monitoring
+
+    - name: Download PDS Overview dashboard
+      get_url:
+        url: https://raw.githubusercontent.com/bluesky-social/pds/main/monitoring/dashboards/pds-overview.json
+        dest: "{{ pds_data_dir }}/monitoring/dashboards/pds-overview.json"
+        mode: "0644"
+      when: pds_telemetry_enabled is defined and pds_telemetry_enabled | bool
+      notify: restart monitoring
+
+    - name: Start monitoring stack
+      shell: cd {{ pds_data_dir }}/monitoring && docker compose up -d
+      when: pds_telemetry_enabled is defined and pds_telemetry_enabled | bool
+
   handlers:
     - name: apply changes
       shell: cd {{ pds_data_dir }} && docker compose up -d && docker restart caddy
+
+    - name: restart monitoring
+      shell: cd {{ pds_data_dir }}/monitoring && docker compose up -d --force-recreate