Lab08 Complete

app_python/app.py

@@ -3,18 +3,50 @@
 Main application module
 """
 import json
-from flask import Flask, jsonify, request
+from flask import Flask, Response, jsonify, request
 from datetime import datetime, timezone
 import logging
 import os
 import platform
 import socket
+from prometheus_client import Counter, Histogram, Gauge, generate_latest
+
 
 HOST = os.getenv('HOST', '0.0.0.0')
 PORT = int(os.getenv('PORT', 5000))
 DEBUG = os.getenv('DEBUG', 'False').lower() == 'true'
 
 
+class PrometheusStats:
+    http_requests_total: Counter
+    http_request_duration_seconds: Histogram
+    http_requests_in_progress: Gauge
+    system_info_duration_seconds: Histogram
+
+    def __init__(self):
+        self.http_requests_total = Counter(
+            'http_requests_total',
+            'Total HTTP requests',
+            ['method', 'endpoint', 'status']
+        )
+        self.http_request_duration_seconds = Histogram(
+            'http_request_duration_seconds',
+            'HTTP request duration',
+            ['method', 'endpoint']
+        )
+        self.http_requests_in_progress = Gauge(
+            'http_requests_in_progress',
+            'HTTP requests currently being processed'
+        )
+        self.system_info_duration_seconds = Histogram(
+            'system_info_duration_seconds',
+            'System stats collection time'
+        )
+
+
+prometheus = PrometheusStats()
+
+
 class JSONFormatter(logging.Formatter):
     def __init__(self):
         super().__init__()
@@ -92,27 +124,33 @@ def get_http_extra_info():
 
 
 @app.route('/')
+@prometheus.http_request_duration_seconds.labels('GET', '/').time()
+@prometheus.http_requests_in_progress.track_inprogress()
 def index():
     """Main endpoint - service and system information."""
     logger.debug(f'Request: {request.method} {request.path}', extra=get_http_extra_info())
-    return jsonify({
-        'service': {
-            'name':        'devops-info-service',
-            'version':     '1.0.0',
-            'description': 'DevOps course info service',
-            'framework':   'Flask'
-        },
-        'system':  get_system_info(),
-        'runtime': get_uptime(),
-        'request': get_request_info(),
-        'endpoints': [
-            {"path": "/",       "method": "GET", "description": "Service information"},
-            {"path": "/health", "method": "GET", "description": "Health check"}
-        ]
-    })
+    with prometheus.system_info_duration_seconds.time():
+        response = {
+            'service': {
+                'name':        'devops-info-service',
+                'version':     '1.0.0',
+                'description': 'DevOps course info service',
+                'framework':   'Flask'
+            },
+            'system':  get_system_info(),
+            'runtime': get_uptime(),
+            'request': get_request_info(),
+            'endpoints': [
+                {"path": "/",       "method": "GET", "description": "Service information"},
+                {"path": "/health", "method": "GET", "description": "Health check"}
+            ]
+        }
+    return jsonify(response)
 
 
 @app.route('/health')
+@prometheus.http_request_duration_seconds.labels('GET', '/health').time()
+@prometheus.http_requests_in_progress.track_inprogress()
 def health():
     logger.debug(f'Request: {request.method} {request.path}', extra=get_http_extra_info())
     return jsonify({
@@ -122,6 +160,13 @@ def health():
     })
 
 
+@app.route('/metrics')
+@prometheus.http_request_duration_seconds.labels('GET', '/metrics').time()
+@prometheus.http_requests_in_progress.track_inprogress()
+def metrics():
+    return Response(response=generate_latest(), status=200, content_type='text/plain')
+
+
 @app.errorhandler(404)
 def notfound_handler(e):
     logger.info('A 404 Not Found error occured', extra=get_http_extra_info())
@@ -140,6 +185,13 @@ def internal_error(e):
     }), 500
 
 
+@app.after_request
+def after_request(response: Response):
+    prometheus.http_requests_total.labels(request.method, request.path,
+                                          str(response.status_code)).inc()
+    return response
+
+
 START_TIME = datetime.now(timezone.utc)
 logger.info('Application starting... Configured with log level=%s', 'DEBUG' if DEBUG else 'INFO')
 if __name__ == '__main__':

app_python/requirements.txt

@@ -1,2 +1,3 @@
 Flask==3.1.0
 gunicorn==24.0.0
+prometheus-client==0.23.1

monitoring/docker-compose.yml

@@ -17,7 +17,7 @@ services:
       resources:
         limits:
           cpus: '1.0'
-          memory: 3G
+          memory: 1G
         reservations:
           cpus: '1'
           memory: 1G
@@ -52,11 +52,8 @@ services:
     deploy:
       resources:
         limits:
-          cpus: '1.0'
-          memory: 1G
-        reservations:
-          cpus: '1'
-          memory: 500M
+          cpus: '0.5'
+          memory: 512M
     healthcheck:
       test: ["CMD-SHELL", "wget --no-verbose --tries=1 --spider http://localhost:3000/api/health || exit 1"]
       interval: 10s
@@ -76,12 +73,37 @@ services:
     deploy:
       resources:
         limits:
-          cpus: '3.0'
-          memory: 4G
-        reservations:
-          cpus: '3'
+          cpus: '0.5'
+          memory: 256M
+    healthcheck:
+      test: ["CMD-SHELL", "curl -f http://localhost:5000/health || exit 1"]
+      interval: 10s
+      timeout: 5s
+      retries: 5
+
+  prometheus:
+    image: prom/prometheus:v3.9.0
+    ports:
+      - 9090:9090
+    volumes:
+      - ${PWD}/prometheus/prometheus.yml:/etc/prometheus/prometheus.yml:ro
+      - prometheus-data:/prometheus:rw
+    command:
+      - '--config.file=/etc/prometheus/prometheus.yml'
+      - '--storage.tsdb.retention.time=15d'
+      - '--storage.tsdb.retention.size=10GB'
+    deploy:
+      resources:
+        limits:
           memory: 1G
+          cpus: '1.0'
+    healthcheck:
+      test: ["CMD-SHELL", "wget --no-verbose --tries=1 --spider http://localhost:9090/-/healthy || exit 1"]
+      interval: 10s
+      timeout: 5s
+      retries: 5
 
 volumes:
   loki-data:
   grafana-data:
+  prometheus-data: