Subscribe to Code to Architecture’s news feed.
Click on “Follow” and decide if you want to get news from Code to Architecture via RSS, as email newsletter, via mobile or on your personal news page.
Subscription to Code to Architecture comes without risk as you can unsubscribe instantly at any time.
You can also filter the feed to your needs via topics and keywords so that you only receive the news from Code to Architecture which you are really interested in. Click on the blue “Filter” button below to get started.
Title: Code to Architecture
Kubernetes resources for JVM services look deceptively simple: pick requests, maybe add limits, and move on.
In practice, these settings define three different realities at once:
what the scheduler believes the pod needs what the Linux kernel enforces under pressure what the JVM thinks it can safely allocate and keep aliveWhen those realities disagree, ...
Autoscaling looks simple on slides: if load rises, add capacity.
In production, it is a set of feedback loops acting on incomplete signals. That is why HPA and VPA do not merely “work together” or “conflict.” They reshape each other’s inputs, timing, and failure modes.
Quick Summary Mechanism What it changes Best fit Common trap HPA replica co...Pod lifecycle mistakes rarely look dramatic in YAML. They show up as rollout stalls, connection resets, stuck draining, traffic going to the wrong pod, or a service that passes liveness but still is not truly ready.
That is why this topic matters. The question is not “should I add a readiness probe?” The real question is whether the platform and the application agree ...
Part 3 is where API evolution stops being a versioning syntax problem and becomes an operations problem.
Most teams know how to create /v2. Far fewer know how to retire /v1, prove clients are ready, and stop “temporary” compatibility code from becoming permanent platform debt.
Quick Summary Question Healthy answer When do we version? only when c...Idempotency keys and dedupe stores for write APIs (Part 3) is not just a diagramming exercise. The hard part is deciding where ownership, failure handling, and change coordination should live once the system is split across services.
Problem 1: Idempotency keys and dedupe stores for write APIs (Part 3)Problem description: We want to use idempotency keys and dedupe...