HPA and VPA: Autoscaling Pods Without Oscillating Through PagerDuty
Horizontal versus vertical autoscaling in Kubernetes, custom metrics, and the cooldown and request-sizing mistakes that cause flapping.
Tag
14 articles · 53 min of reading
Horizontal versus vertical autoscaling in Kubernetes, custom metrics, and the cooldown and request-sizing mistakes that cause flapping.
sync-wave annotations, hooks, and waves that wait: how to roll CRDs before CRs, and the deadlock of a wave that never becomes healthy.
Why Shopify migrated its core commerce platform out of self-managed data centers onto Google Cloud and Kubernetes, and how it did so without downtime.
As order volume climbed, DoorDash treated infrastructure cost as an engineering metric teams could see and act on, not just a line on a finance report.
Why Stripe moved its compute platform onto Kubernetes, and the engineering work required to run a payments company's workloads on top of it safely.
Why naive CPU-based autoscaling fails under real traffic patterns, and the metric choices and tuning that make autoscaling respond correctly instead of thrashing.
Inside Borg, Google's internal cluster manager, and how its ideas about scheduling, priority, and bin packing became the design for Kubernetes.
EKS, GKE, and AKS all promise to take Kubernetes operations off your plate, but the actual division of responsibility is narrower than most teams assume.
A concrete comparison of blue-green and canary deployment strategies, including the infrastructure they each require and where one clearly beats the other.
How Pinterest migrated thousands of services onto Kubernetes and built internal platform abstractions so product teams didn't need to become infrastructure experts.
A comparison of secrets management patterns, from environment variables to dedicated vaults, and how to pick the right one without overengineering a simple app.
How GitOps replaces imperative deploy scripts with a reconciled, auditable source of truth, and the practical patterns for running it well with Argo CD or Flux.
A practical look at how Helm's values and templating actually resolve, plus the recurring mistakes that turn a simple chart into a debugging session.
The difference between liveness, readiness, and startup probes in Kubernetes, and the common misconfigurations that cause restart loops and dropped traffic.