Black Friday sale: an extra 40% off every course until Monday, November 30.New Year sale: an extra 30% off every course until Friday, January 1.Memorial Day sale: an extra 25% off every course until Monday, May 31.SysAdmin Day sale: an extra 25% off every course until Sunday, August 1.Back to School sale: an extra 25% off every course until Sunday, August 22.Labor Day sale: an extra 25% off every course until Monday, September 6.Black Friday sale: an extra 40% off every course until Monday, November 29.New Year sale: an extra 30% off every course until Saturday, January 1.Choose your certification →

Copyright (c) 2026 MindMesh Academy. All rights reserved. This content is proprietary and may not be reproduced or distributed without permission.

3.5.1.7. Container Monitoring, Multi-Account Log Aggregation, and Alerting (Container Insights, Log Subscriptions, CloudWatch Alarms)

💡 First Principle: Collect telemetry at the level where failures actually happen (container and pod, not just the host), move it to one place without per-account plumbing, and alert on conditions a human must act on, not on every metric that moved.

Monitoring containerized applications:
  • "CloudWatch Container Insights" collects metrics and performance log events (embedded metric format) for "ECS", "EKS", ROSA and Kubernetes on EC2, including tasks on "Fargate". It aggregates at cluster, node, pod, task and service level and surfaces diagnostics such as container restart failures.
  • Container Insights with enhanced observability adds container-level detail and curated dashboards. On ECS it is a setting: containerInsights=enhanced as the account default or per cluster (enabled gives the original version). Instance-level metrics for EC2-backed ECS clusters need the "CloudWatch agent" as a daemon. On EKS install the Amazon CloudWatch Observability EKS add-on, which deploys the CloudWatch agent and Fluent Bit and turns on Container Insights and Application Signals by default. Enhanced observability for EKS is billed per observation rather than per custom metric.
  • "AWS Distro for OpenTelemetry (ADOT)" is the AWS-supported OpenTelemetry collector and SDKs: one vendor-neutral instrumentation that can export metrics to "CloudWatch" or "Amazon Managed Service for Prometheus" and traces to "X-Ray". Choose it when teams already use Prometheus or OpenTelemetry, or must keep the option of another backend.
  • Container logs: Fluent Bit (EKS) or FireLens (ECS) route stdout/stderr to CloudWatch Logs or Firehose.
Multi-account log aggregation and archives:
  • Subscription filters stream log events in real time to "Kinesis Data Streams", "Firehose" or "Lambda". Cross-account delivery goes through a CloudWatch Logs destination in the central account (same Region as the source log group) whose access policy names the sender accounts or the organization. An account-level subscription filter policy applies one filter to all log groups in an account, so new log groups are covered without per-group setup.
  • Firehose to S3 in a log-archive account gives the long-term store: S3 Lifecycle moves objects to S3 Glacier Instant Retrieval, Flexible Retrieval or Deep Archive by age (Deep Archive is the cheapest tier; standard retrievals complete within about 12 hours), Object Lock adds WORM retention (compliance mode: nobody, including the root user, can delete or shorten retention; governance mode: principals with s3:BypassGovernanceRetention can), and Athena queries the archive. Set a short retention on the source log groups so you do not pay twice.
  • CloudWatch Logs centralization rules (configured from the management or delegated administrator account) copy new log events from chosen accounts, OUs and Regions into log groups in a destination account, with an optional backup Region. They copy only data that arrives after the rule exists, so they do not backfill history.
  • CloudWatch cross-account observability (OAM, see 3.5.1) lets a monitoring account query source accounts' logs and metrics in place; it is a viewing model, not an archive.
Alerting design:
  • Alarm types: static threshold, anomaly detection bands for metrics with daily or weekly seasonality, Metrics Insights alarms that watch a whole fleet with one query, and composite alarms that combine child alarms with AND/OR/NOT (for example "error rate high AND latency high") to cut alert noise. Composite alarms cannot perform Auto Scaling actions.
  • Actions: "SNS" (email, SMS, chat, HTTPS endpoints), direct "Lambda" invocation, EC2 stop, terminate, reboot or recover, Auto Scaling, and OpsItems in "Systems Manager OpsCenter" for tracked human follow-up. Every state change is also an "EventBridge" event, which can start an SSM Automation runbook. Alarm mute rules silence actions during planned maintenance windows while monitoring continues.
  • Use M out of N datapoints and TreatMissingData deliberately: a missing heartbeat metric should usually be treated as breaching.

⚠️ Exam Trap: "AWS Systems Manager Incident Manager" stopped accepting new customers on November 7, 2025. For a new design, route alarms to SNS/EventBridge, track work in OpsCenter, and use a partner on-call tool for paging; do not choose Incident Manager response plans as the answer for a new account.

See how it connects
Alvin Varughese
Written byAlvin Varughese
Founder•26 professional certifications earned