640 posts tagged with "Kubernetes"
Container orchestration, Cluster API, and self-hosted control planes
Scale Before the Spike: Predictive GPU Pre-Warming for a Cluster API Fleet Without the Always-On Bill
Reactive autoscaling leaves a roughly 45-minute gap between an inference spike and ready GPUs. A concrete pre-warm design for a Cluster API fleet — request-side forecast signals, a measured boot budget, and a capped warm reserve — priced against Hetzner's monthly GPU boxes.
SOPS with Age vs Sealed Secrets: What GitOps Secrets Cost to Rotate on a Cluster API Fleet
Committing encrypted secrets to Git is the easy part. A fleet-scale comparison of SOPS with age versus Bitnami Sealed Secrets on what each costs to rotate, audit, and recover after the management cluster dies — with runbooks and the February 2026 rotation CVE that settles it.
Kubernetes 1.37's StatefulSet Recreate Strategy (Alpha): Stop Deleting Stuck Pods by Hand
Kubernetes 1.37 adds an alpha Recreate update strategy for StatefulSets that deletes old-revision pods wholesale instead of halting behind the first stuck pod. What it changes, the full-outage price, and where a self-hosted PaaS may use it — and where it must not.
Your Three Replicas Are All in FSN1: TopologySpreadConstraints, Hetzner Failure Domains, and the Placement Policy a Self-Hosted PaaS Owes Its Tenants
Two replicas across three Hetzner locations at maxSkew 1 lands 1/1/0, with one location always empty. The skew formula behind it, the minDomains trap that wedges pods in Pending, what cross-location spread really costs on Hetzner's private networks, and the soft-by-default placement policy a self-hosted PaaS should ship.
Don't Build Cron Into Your PaaS: Running Windmill's Script-to-Webhook Engine as a Tenant Workload
Every git-push PaaS eventually needs cron and background workers. A concrete build-vs-route decision: run Windmill's open-source workflow engine as a tenant workload, or build your own — with the six-dimension tradeoff table and the three conditions that flip the answer.
When the Grid Itself Is the Threat Model: What Berlin's Four-Day Blackout Teaches About Fleet Placement
A January 2026 arson attack naming AI data centers left 45,000 Berlin households dark for four days. Why concentrated compute is now an explicit target — and why a distributed self-hosted fleet is the cheapest hedge.
Karpenter Still Doesn't Run on Bare Metal: What vCluster's Auto Nodes Workaround Really Buys You
Karpenter provisions right-sized cloud nodes in under a minute, but it still has no bare-metal provider in 2026. A stage-by-stage latency budget shows what vCluster's Terraform-backed Auto Nodes actually recovers — and why Cluster Autoscaler remains the sane default for a Cluster API fleet on owned hardware.
KEDA vs Knative vs Sablier: Three Ways to Make Idle Kubernetes Services Cost Nothing
KEDA, Knative, and Sablier put the scale-to-zero interception layer in three different places. A concrete comparison of cold starts, config cost, and probe hygiene — plus the health-check audit that decides whether any of them saves money.
Longhorn V2's Kernel Bypass: What SPDK Storage Actually Buys Your Self-Hosted Volumes
Longhorn's V2 engine moves block storage into userspace with SPDK for lower latency than the iSCSI-based V1 path — at the cost of a CPU core, 2 GiB of hugepages, and raw block disks per node. Measured gains, hardware caveats, and when to pilot it on your own metal.