---
# Comprehensive Gitea Timeout Diagnosis
# Prüft alle Aspekte des intermittierenden Gitea-Timeout-Problems
- name: Comprehensive Gitea Timeout Diagnosis
  hosts: production
  gather_facts: yes
  become: no
  vars:
    gitea_stack_path: "{{ stacks_base_path }}/gitea"
    traefik_stack_path: "{{ stacks_base_path }}/traefik"
    gitea_url: "https://{{ gitea_domain }}"

  tasks:
    - name: Check Traefik container uptime and restart count
      ansible.builtin.shell: |
        docker inspect traefik --format '{{ '{{' }}.State.Status{{ '}}' }}|{{ '{{' }}.State.StartedAt{{ '}}' }}|{{ '{{' }}.RestartCount{{ '}}' }}' 2>/dev/null || echo "UNKNOWN"
      register: traefik_info
      changed_when: false

    - name: Check Gitea container uptime and restart count
      ansible.builtin.shell: |
        docker inspect gitea --format '{{ '{{' }}.State.Status{{ '}}' }}|{{ '{{' }}.State.StartedAt{{ '}}' }}|{{ '{{' }}.RestartCount{{ '}}' }}' 2>/dev/null || echo "UNKNOWN"
      register: gitea_info
      changed_when: false

    - name: Check Traefik logs for recent restarts (last 2 hours)
      ansible.builtin.shell: |
        cd {{ traefik_stack_path }}
        docker compose logs traefik --since 2h 2>&1 | grep -iE "stopping server gracefully|I have to go|restart|shutdown" | tail -20 || echo "Keine Restart-Meldungen in den letzten 2 Stunden"
      register: traefik_restart_logs
      changed_when: false

    - name: Check Gitea logs for errors/timeouts (last 2 hours)
      ansible.builtin.shell: |
        cd {{ gitea_stack_path }}
        docker compose logs gitea --since 2h 2>&1 | grep -iE "error|timeout|failed|panic|fatal|slow" | tail -30 || echo "Keine Fehler in den letzten 2 Stunden"
      register: gitea_error_logs
      changed_when: false

    - name: Test Gitea direct connection (multiple attempts)
      ansible.builtin.shell: |
        for i in {1..5}; do
          echo "=== Attempt $i ==="
          cd {{ gitea_stack_path }}
          timeout 5 docker compose exec -T gitea curl -f http://localhost:3000/api/healthz 2>&1 || echo "FAILED"
          sleep 1
        done
      register: gitea_direct_tests
      changed_when: false

    - name: Test Gitea via Traefik (multiple attempts)
      ansible.builtin.shell: |
        for i in {1..5}; do
          echo "=== Attempt $i ==="
          timeout 10 curl -k -s -o /dev/null -w "%{http_code}" {{ gitea_url }}/api/healthz 2>&1 || echo "TIMEOUT"
          sleep 2
        done
      register: gitea_traefik_tests
      changed_when: false

    - name: Check Traefik service discovery for Gitea (using CLI)
      ansible.builtin.shell: |
        cd {{ traefik_stack_path }}
        docker compose exec -T traefik traefik show providers docker 2>/dev/null | grep -i "gitea" || echo "Gitea service not found in Traefik providers"
      register: traefik_gitea_service
      changed_when: false
      failed_when: false

    - name: Check Traefik routers for Gitea (using CLI)
      ansible.builtin.shell: |
        cd {{ traefik_stack_path }}
        docker compose exec -T traefik traefik show providers docker 2>/dev/null | grep -i "gitea" || echo "Gitea router not found in Traefik providers"
      register: traefik_gitea_router
      changed_when: false
      failed_when: false

    - name: Check network connectivity Traefik -> Gitea
      ansible.builtin.shell: |
        cd {{ traefik_stack_path }}
        for i in {1..3}; do
          echo "=== Attempt $i ==="
          docker compose exec -T traefik wget -qO- --timeout=5 http://gitea:3000/api/healthz 2>&1 || echo "CONNECTION_FAILED"
          sleep 1
        done
      register: traefik_gitea_network
      changed_when: false

    - name: Check Gitea container resources (CPU/Memory)
      ansible.builtin.shell: |
        docker stats gitea --no-stream --format 'CPU: {{ '{{' }}.CPUPerc{{ '}}' }} | Memory: {{ '{{' }}.MemUsage{{ '}}' }}' 2>/dev/null || echo "Could not get stats"
      register: gitea_resources
      changed_when: false
      failed_when: false

    - name: Check Traefik container resources (CPU/Memory)
      ansible.builtin.shell: |
        docker stats traefik --no-stream --format 'CPU: {{ '{{' }}.CPUPerc{{ '}}' }} | Memory: {{ '{{' }}.MemUsage{{ '}}' }}' 2>/dev/null || echo "Could not get stats"
      register: traefik_resources
      changed_when: false
      failed_when: false

    - name: Check if Gitea is in traefik-public network
      ansible.builtin.shell: |
        docker network inspect traefik-public --format '{{ '{{' }}range .Containers{{ '}}' }}{{ '{{' }}.Name{{ '}}' }} {{ '{{' }}end{{ '}}' }}' 2>/dev/null | grep -q gitea && echo "YES" || echo "NO"
      register: gitea_in_network
      changed_when: false

    - name: Check Traefik access logs for Gitea requests (last 100 lines)
      ansible.builtin.shell: |
        cd {{ traefik_stack_path }}
        tail -100 logs/access.log 2>/dev/null | grep -i "git.michaelschiemer.de" | tail -20 || echo "Keine Access-Logs gefunden"
      register: traefik_access_logs
      changed_when: false
      failed_when: false

    - name: Check Traefik error logs for Gitea-related errors
      ansible.builtin.shell: |
        cd {{ traefik_stack_path }}
        tail -100 logs/traefik.log 2>/dev/null | grep -iE "gitea|git\.michaelschiemer\.de|timeout|error.*gitea" | tail -20 || echo "Keine Gitea-Fehler in Traefik-Logs"
      register: traefik_error_logs
      changed_when: false
      failed_when: false

    - name: Summary
      ansible.builtin.debug:
        msg: |
          ================================================================================
          UMFASSENDE GITEA TIMEOUT DIAGNOSE:
          ================================================================================
          
          Container Status:
          - Traefik: {{ traefik_info.stdout }}
          - Gitea: {{ gitea_info.stdout }}
          
          Traefik Restart-Logs (letzte 2h):
          {{ traefik_restart_logs.stdout }}
          
          Gitea Error-Logs (letzte 2h):
          {{ gitea_error_logs.stdout }}
          
          Direkte Gitea-Verbindung (5 Versuche):
          {{ gitea_direct_tests.stdout }}
          
          Gitea via Traefik (5 Versuche):
          {{ gitea_traefik_tests.stdout }}
          
          Traefik Service Discovery:
          - Gitea Service: {{ traefik_gitea_service.stdout }}
          - Gitea Router: {{ traefik_gitea_router.stdout }}
          
          Netzwerk-Verbindung Traefik -> Gitea (3 Versuche):
          {{ traefik_gitea_network.stdout }}
          
          Container-Ressourcen:
          - Gitea: {{ gitea_resources.stdout }}
          - Traefik: {{ traefik_resources.stdout }}
          
          Netzwerk:
          - Gitea in traefik-public: {% if gitea_in_network.stdout == 'YES' %}✅{% else %}❌{% endif %}
          
          Traefik Access-Logs (letzte 20 Gitea-Requests):
          {{ traefik_access_logs.stdout }}
          
          Traefik Error-Logs (Gitea-bezogen):
          {{ traefik_error_logs.stdout }}
          
          ================================================================================
          ANALYSE:
          ================================================================================
          
          {% if 'stopping server gracefully' in traefik_restart_logs.stdout | lower or 'I have to go' in traefik_restart_logs.stdout %}
          ❌ PROBLEM: Traefik wird regelmäßig gestoppt!
          → Dies ist die Hauptursache für die Timeouts
          → Führe 'find-traefik-restart-source.yml' aus um die Quelle zu finden
          {% endif %}
          
          {% if 'CONNECTION_FAILED' in traefik_gitea_network.stdout %}
          ❌ PROBLEM: Traefik kann Gitea nicht erreichen
          → Netzwerk-Problem zwischen Traefik und Gitea
          → Prüfe ob beide Container im traefik-public Netzwerk sind
          {% endif %}
          
          {% if 'not found' in traefik_gitea_service.stdout | lower or 'not found' in traefik_gitea_router.stdout | lower %}
          ❌ PROBLEM: Gitea nicht in Traefik Service Discovery
          → Traefik hat Gitea nicht erkannt
          → Führe 'fix-gitea-timeouts.yml' aus um beide zu restarten
          {% endif %}
          
          {% if 'TIMEOUT' in gitea_traefik_tests.stdout %}
          ⚠️  PROBLEM: Intermittierende Timeouts via Traefik
          → Mögliche Ursachen: Traefik-Restarts, Gitea-Performance, Netzwerk-Probleme
          {% endif %}
          
          ================================================================================