Cloud
AWS Lambda Failure Modes You Should Budget For
AWS Lambda Failure Modes You Should Budget For
Timeouts, retries, partial batch failure, and the handler that is not idempotent.
Cloud
Lambda, queues, IAM, containers, and the operational details that turn diagrams into systems.
AWS · AWS Lambda · Docker · 5 published · 19 in the review queue. Written by Al Beltran.
Cloud
AWS Lambda Failure Modes You Should Budget For
Timeouts, retries, partial batch failure, and the handler that is not idempotent.

Hybrid key exchange is landing in browsers and CDNs. The engineering work is inventory, dual-stack certificates, and not breaking the middleboxes you forgot you had.

Managed functions removed capacity planning. They did not remove tail latency, cold starts, or the cost of chatty downstreams. You still have a budget. It is just itemized differently.

xz and a decade of npm incidents taught the same lesson: you run other people's code. Pinning, provenance, and a small dependency graph beat a PDF named policy.

What matters when Lambda, S3, and Step Functions process high-volume data under production SLAs.
Planned notes stay unpublished until they are written and reviewed. Titles are listed so the topic map is complete. They are not pages.
AWS Interview Questions, With the Reasoning
IAM, failure modes, and the services you should name only if you can describe the blast radius.
interview · mixed · high priority
Docker Interview Questions, With the Reasoning
Images, layers, processes, and the difference between "it runs on my machine" and a production image.
interview · mixed · high priority
IAM Least Privilege in Practice
Start from the action the function takes. "AdministratorAccess for now" becomes forever.
educational · intermediate · medium priority
SQS Visibility Timeout vs the Work Duration
If processing is longer than visibility, you will double-process. If it is too long, failures sit invisible.
troubleshooting · intermediate · medium priority
S3 Prefixes, Listing, and the Consistency You Can Rely On
What listing costs, how keys should be designed, and what "read after write" does not mean for every operation.
informational · intermediate · medium priority
VPC ENIs and Lambda Cold Starts
When the function needs a VPC, what you pay in init, and the alternatives worth checking first.
troubleshooting · intermediate · medium priority
CloudWatch Alarms That Page a Human
Burn rate, missing data, and the alarm that fired because a batch job is allowed to fail.
educational · intermediate · medium priority
Docker Multi-Stage Builds That Keep Secrets Out of Layers
Build deps vs runtime deps, and the COPY that leaked a .env.
educational · intermediate · medium priority
ECS vs Lambda: Pick from the Latency and Ops Budget
Always-on vs burst. Chatty connections vs request-scoped compute.
comparison · intermediate · medium priority
When Step Functions Are the Orchestrator (and When They Are a Tax)
Long workflows, retries, and the state machine that should have been a queue.
comparison · intermediate · medium priority
Secrets Manager vs Environment Variables
Rotation, blast radius, and the Lambda env that still shows up in a screenshot.
comparison · intermediate · medium priority
CI/CD for a Static Next.js Site Without Drama
Build, preview, promote. Keep secrets out of the client bundle.
educational · intermediate · medium priority
Terraform State and Blast Radius
Who can apply, what a workspace contains, and the destroy you will not run from a laptop.
architecture · senior · medium priority
AWS Cost: The Query, the Log, and the NAT You Forgot
Chatty Lambdas, chatty logs, and data transfer that does not show up on the architecture slide.
problem-solving · intermediate · medium priority
Blue/Green vs Feature Flags
Infrastructure cutover vs in-process rollout. You often want both.
comparison · intermediate · medium priority
Container Image Supply Chain: Base, Pin, Scan
Digest pins, SBOM habits, and the latest tag that is not a version.
educational · intermediate · medium priority
API Gateway Timeouts vs Lambda Timeouts
The shorter timeout wins. The user sees 504 while your function is still writing.
troubleshooting · intermediate · medium priority
The Three Signals: Latency, Errors, Saturation
If your dashboard cannot say whether to wake someone, it is not an SLO dashboard.
informational · intermediate · medium priority
Backups You Have Restored at Least Once
A backup that has never been restored is a hope. Rehearse the restore.
educational · intermediate · medium priority