Skip to content

AWS ECS Fargate Magento 2.4.9 live cell — 2026-08-13 (20260813ai)

This is a bounded Magento application cold baseline on AWS ECS Fargate. It proves signed-digest promote, seed import into RDS, ECS runtime health, a warm queueMode:db infra-only health check, destroy of the 101-resource stack, disposable state-bucket deletion, prerequisite-secret force-delete, and independent empty stack inventories. It does not claim ALB HTTP 200, custom-domain TLS, Magento origin through CloudFront or Fastly, Fargate Spot/EC2/Managed Instances interruption, HA, DR, collectors, or WAF.

Earlier attempt 20260813ah failed at magelift promote (signed release verification failed) with created=0. That run ID is not a PASS.

Scope

Field Value
Account <aws-account>
Region eu-west-3
Run ID 20260813ai
Project tag awsah
Profile preview
Runtime ecs-fargate (computeMode:fargate)
Catalog spec certification-aws
Cell runtime ecs-fargate, compute fargate, Magento 2.4.9, rds-mysql 8.4.10, search disabled, queue db, digest below
Network single-AZ fck-nat, CIDR 10.83.0.0/16
Edge edge.mode: none (media CloudFront only)
Catalog scripts/acceptance/cells-aws-ecs-architecture-fargate.txt
Database RDS MySQL 8.4.10, database magento
Cache ElastiCache Valkey 9.0 cache.t4g.micro
Queue queueMode:db
Search disabled
Artifact ECR magelift-acceptance-rc1-20260813@sha256:e82dcf767cf57492b1c5619eac708d9af8b2e66f7b144ff49e6b2513753db495
Cosign devops@<gcp-project>.iam.gserviceaccount.com / https://accounts.google.com
Seed .magelift/seed/magento-249-sanitized-definer-free.sql.gz (gitignored; values not logged)
Wrapper scripts/aws-acceptance-local.sh
Wall clock 3364723 ms (~56 min), exit 0

Destroyed identities (not live): VPC vpc-0b19091fa6feba3e7, ALB awsah-preview-ingress-alb-1981458961.eu-west-3.elb.amazonaws.com, media CloudFront droyr6v2qdokt.cloudfront.net, RDS awsah-preview-database-instance6e11f90.cvcuaioecipe.eu-west-3.rds.amazonaws.com, ECS cluster awsah-preview-runtime-cluster-984e6ca, stack awsah-preview-aws-ecs-fargate.

Observed result

  1. Promote verified the Cosign signature on the ECR digest. Create-once planned 101 resources. Seed import completed (database=magento).
  2. Baseline computeMode:fargate PASSed in 1982s. Runtime health was ECS-only: service 1/1 running, primary rollout completed, task healthy. Catalog has no cutover:dns. Custom-domain HTTPS was not exercised.
  3. Warm queueMode:db used --infra-only, retried while rollout was incomplete, then reached the same ECS runtime healthy state (result=PASS).
  4. EXIT destroy deleted 101 resources in 19m42s, then deleted state bucket magelift-<aws-account>-eu-west-3-awsah-preview-state, RDS error log group, and force-deleted the three awsah prerequisite secrets. Harness assert_clean ok (tag index still reported 38 mappings; live inventories passed).

Independent inventory after destroy

Empty: VPC, RDS, ECS clusters, ALB named awsah, ElastiCache, S3 named awsah, Secrets Manager named awsah, CloudFront distributions.

Immediately after destroy: unused ISSUED ACM in eu-west-3/us-east-1 and two Cloudflare ACM-validation CNAMEs (bootstrap from 20260813ah, InUse=[]). Follow-up deleted those CNAMEs (marker magelift/acceptance/acm-dns/20260813ah) and both certificates (ResourceNotFoundException on describe).

Still recorded, not stack resources:

  • Customer KMS 237f4553-5cb1-4c15-ace5-824dc44e2301 Enabled (do not delete)
  • EC2 i-<redacted> State=terminated (fck-nat tombstone)

Non-claims

Spot/EC2/Managed Instances live interruption, EKS, ALB HTTP through the hostname, Magento Admin/storefront HTTPS on ml-aws-20260813ah.example.test, Magento origin on CloudFront/Fastly, HA, DR, fixtures, and collector wiring remain open.