Pular para conteúdo

CloudWatch

Logs e métricas com Amazon CloudWatch.

Logs

Structured Logging (JSON)

Python (Backend):

import json
import logging
from pythonjsonlogger import jsonlogger

logger = logging.getLogger()
logHandler = logging.StreamHandler()
formatter = jsonlogger.JsonFormatter()
logHandler.setFormatter(formatter)
logger.addHandler(logHandler)
logger.setLevel(logging.INFO)

# Usage
logger.info("User logged in", extra={
    "user_id": user.id,
    "email": user.email,
    "ip_address": request.client.host,
    "correlation_id": correlation_id
})

Output:

{
  "message": "User logged in",
  "user_id": 123,
  "email": "user@example.com",
  "ip_address": "192.168.1.1",
  "correlation_id": "abc-123-def",
  "timestamp": "2026-01-20T10:30:00Z",
  "level": "INFO",
  "logger": "app.auth"
}

Log Groups

Cada Lambda tem seu log group:

  • /aws/lambda/app-user-api-staging
  • /aws/lambda/app-user-api-production
  • /aws/lambda/app-payment-processor-staging
  • /aws/lambda/app-payment-processor-production

Retention

  • Staging: 30 dias
  • Production: 90 dias

Searching Logs

# Via AWS CLI
aws logs filter-log-events \
  --log-group-name /aws/lambda/app-user-api-production \
  --filter-pattern '{ $.user_id = 123 }' \
  --start-time $(date -d '1 hour ago' +%s)000

# Via CloudWatch Insights

CloudWatch Insights Queries:

-- Erros na última hora
fields @timestamp, @message, error
| filter level = "ERROR"
| sort @timestamp desc
| limit 100

-- Requests lentos (> 1s)
fields @timestamp, user_id, duration
| filter duration > 1000
| stats avg(duration), max(duration), count() by bin(1m)

-- Requests por usuário
fields user_id
| stats count() by user_id
| sort count() desc

Correlation IDs

Para rastrear requests através de múltiplos serviços:

import uuid
from contextvars import ContextVar

correlation_id_var: ContextVar[str] = ContextVar('correlation_id')

# Middleware FastAPI
@app.middleware("http")
async def add_correlation_id(request: Request, call_next):
    correlation_id = request.headers.get('X-Correlation-ID') or str(uuid.uuid4())
    correlation_id_var.set(correlation_id)

    response = await call_next(request)
    response.headers['X-Correlation-ID'] = correlation_id
    return response

# Logger com correlation ID
def get_logger():
    correlation_id = correlation_id_var.get(None)
    return logger.bind(correlation_id=correlation_id)

Métricas

Lambda Metrics

AWS Managed: - Invocations - Número de invocações - Errors - Número de erros - Duration - Duração (ms) - Throttles - Throttling events - ConcurrentExecutions - Execuções simultâneas - IteratorAge - Idade do batch (SQS/Kinesis)

Custom Metrics:

import boto3

cloudwatch = boto3.client('cloudwatch')

def publish_metric(metric_name: str, value: float, unit: str = 'Count'):
    """Publish custom metric to CloudWatch."""
    cloudwatch.put_metric_data(
        Namespace='MyApp',
        MetricData=[{
            'MetricName': metric_name,
            'Value': value,
            'Unit': unit,
            'Timestamp': datetime.utcnow()
        }]
    )

# Usage
publish_metric('UserSignups', 1, 'Count')
publish_metric('PaymentAmount', 99.99, 'None')
publish_metric('APILatency', 250, 'Milliseconds')

SQS Metrics

  • NumberOfMessagesSent
  • NumberOfMessagesReceived
  • ApproximateAgeOfOldestMessage - Importantíssimo!
  • ApproximateNumberOfMessagesVisible

RDS Metrics

  • DatabaseConnections
  • CPUUtilization
  • FreeableMemory
  • ReadLatency / WriteLatency
  • FreeStorageSpace

Dashboards

Dashboard Principal

Widgets:

  1. API Health
  2. Request count (last hour)
  3. Error rate

  4. Average latency (p50, p90, p99)

  5. Lambda Performance

  6. Invocations por função
  7. Duration metrics
  8. Error count

  9. Throttles

  10. Queue Health

  11. Messages in queue
  12. Age of oldest message

  13. DLQ messages

  14. Database

  15. Connections
  16. CPU usage

  17. Disk space

  18. Business Metrics

  19. User signups (today)
  20. Orders created (today)
  21. Revenue (today)

Criar Dashboard

# Via AWS CLI
aws cloudwatch put-dashboard \
  --dashboard-name "App-Production" \
  --dashboard-body file://dashboard.json

Referências