Writing
Notes on reliability, cloud cost and AWS by Abhinav C.
- Terraform state is a production databaseIt holds every resource ID, sometimes secrets, and the only record of what Terraform thinks it owns. Treat it with the same care as the database it probably describes.
- Ingress NGINX is retired. Your ingress still works. For now.The community ingress-nginx controller stopped getting fixes in March 2026. Nothing broke that day, which is exactly why it is easy to ignore.
- Certificates now last 200 days. Then 47.Public TLS certificates are on a schedule to shrink to 47 days by 2029. If renewal still involves a person, the schedule is the incident.
- Jenkins isn't your problem. Your Jenkinsfile is.Teams migrate off Jenkins to escape slow, fragile pipelines, then rebuild the same pipelines somewhere else. The tool was never the bottleneck. The design was.
- gflow: team conventions belong in a config file, not a wiki pageEvery team has rules for branch names, commit messages and pull requests. Most keep them in a document people skim once. gflow puts them in a committed YAML file that a CLI follows for you.
- Measure Kafka lag in seconds, not messagesA consumer group 50,000 messages behind might be fine or might be an hour stale. The number that matters to the business is how old the data is.
- Moving EC2 to ECS without anyone noticingThe best migration is one where the only people who notice are the ones reading the bill.
- Scheduling Kubernetes jobs for when the grid is cleanerA carbon-aware scheduler that delays batch jobs until electricity is cleaner. The interesting decisions were not about carbon at all. They were about what happens when the data is missing.