Amazon CloudWatch insights

CloudWatch agent configuration changes don't take effect because agent was restarted without fetch-config option. Agent continues using old cached configuration despite updated config file.

CloudWatch agent running as container service fails or takes excessive time to start due to IMDSv2 hop limit restrictions when accessing EC2 metadata. Agent logs 'Please increase hop limit to 2' errors.

CloudWatch agent shows 'running' status but fails to send metrics due to IAM permission issues, network connectivity problems, or configuration errors. Agent status checks alone are insufficient to confirm data delivery.

CloudWatch agent in private subnets fails to send metrics when VPC endpoints are misconfigured or security groups block traffic. Agents report network timeout errors despite appearing to have network connectivity.

CloudWatch agent publishes metrics successfully but they appear 'missing' because users look in wrong namespace. Default namespace is 'CWAgent', not 'AWS/EC2', causing confusion when metrics don't appear in expected locations.

ECS Service Degradation Hidden by Averages

Default CloudWatch metrics for ECS show service-wide averages that mask individual task problems. A service may appear healthy at 60% average memory while some tasks are at 95% and approaching OOM kills.

Missing CloudWatch Alarms for Infrastructure Provisioned by Terraform

Critical infrastructure components (Lambda, ECS, SQS, S3) provisioned via Terraform lack corresponding CloudWatch alarms, resulting in silent failures and extended mean-time-to-detection for production issues.

CloudWatch Application Signals captures technical RED metrics (Rate, Errors, Duration) but lacks business context. Teams can see latency increased but not how many high-value transactions failed or revenue impact of degradation.

High-Value Transaction Timeout Pattern

Enterprise transactions exceeding $10,000 experience significantly higher latency due to enhanced fraud detection, causing timeouts and abandoned purchases. Standard monitoring shows healthy services while high-value orders fail.

Inventory Capacity Masquerading as Performance Issue

ECS inventory service appears healthy with passing health checks and normal container metrics, but business metrics reveal only 15% stock availability due to insufficient capacity for business volume, causing lost sales.

EC2 notification service shows healthy instance metrics but delivers only 171 notifications despite thousands of triggering events. Queue metrics reveal 6,000+ message backlog invisible to standard infrastructure monitoring.

By comparing CloudWatch's SuccessfulRequestLatency metric (service-side processing time) with client-measured end-to-end latency, operators can isolate network overhead from service-level delays and target optimization efforts appropriately.

Organizations running AWS workloads across multiple accounts lack unified visibility, causing delayed incident response. Without centralized observability, teams must check multiple consoles to understand issues spanning accounts.

Excessive CloudWatch alarms on non-actionable metrics cause alert fatigue, leading teams to ignore or miss critical notifications. Alerts fire frequently on normal behavior variations rather than actual service degradation.