Dora Noda
Software Engineer
2333 posts · View all authors
etcd v3.7's RangeStream Fixes a Decade-Old List-Watch Bug — But Not the One That Actually Broke Cluster API at Scale
etcd v3.7 ships RangeStream, a real fix for a decade-old memory-buffering bug in large list-watch reads. But Cluster API's own 300-cluster scale test shows the actual ceiling on a management cluster's etcd comes from somewhere else entirely.
Self-Hosted Kubernetes Installs Don't Stall on Kubernetes — They Stall on Your Registry, Network, and Approval Chain
Fairwinds' CTO named the real reason self-hosted Kubernetes installs drag on for months — and it's rarely Kubernetes. Here's what that means for any Cluster-API PaaS that might one day sell into a customer's own infrastructure.
Fly.io Kills GPUs on August 1, 2026: Where Do Your Agent Sandboxes Go Now?
Fly.io fully deprecates GPU Machines on July 31, 2026. Here's the duty-cycle cost breakdown across RunPod, Modal, and a self-hosted Hetzner GPU box for the teams whose agent sandboxes and fine-tuning jobs have to move.
Gartner's 80%-by-2026 Platform Team Number Is Real. Here's the Golden-Path Math That Decides If It Was Worth It
Gartner's 80%-of-orgs-by-2026 platform-team prediction is on track, but the number that actually predicts a lead-time win is the golden path underneath it: 80%+ voluntary adoption with one, under 20% without.
Headlamp's Cluster API Plugin: Browser-Native Fleet Management Arrives for Clusters You Actually Own
Headlamp's new Cluster API plugin puts MachineDeployment and MachineSet reconciliation state in a browser tab instead of a kubectl session — a concrete before/after for anyone running a self-hosted fleet on owned hardware.
Hetzner's June 2026 Price Hike Doubled Some Instances — Here's the Recomputed Break-Even Against Render and Railway
Hetzner's June 15, 2026 adjustment raised CPX and CCX instances by up to 209% while CX and CAX barely moved. A line-by-line recompute of a web+worker+Postgres topology on post-hike Hetzner versus Render and Railway — and why picking the wrong instance family now costs as much as picking the wrong vendor.
ingress-nginx's EOL Didn't Stop CVE-2026-3288: What Patching a Dead Project Means for Your Routing Layer
ingress-nginx's official end of life didn't stop three RCE patches from landing in its final weeks and months after — here's the sourced timeline, a fleet-wide audit you can run today, and why the blast radius is worse on a self-hosted PaaS.
JADEPUFFER Ran a Full Ransomware Kill Chain With No Human Involved — What That Means for Agents That Deploy Your App
An AI agent ran an entire ransomware attack end to end with no human directing a single step. Here's what that kill chain means for any platform whose own agent can deploy or restart your infrastructure.
Karpenter's Alpha Cluster API Provider vs. Cluster Autoscaler: What Actually Autoscales an Owned-Hardware Kubernetes Fleet in 2026
A Hetzner-backed Cluster API fleet gets its node-ready time from Cluster API's own machine lifecycle, not from which autoscaler is asking — here's what Cluster Autoscaler, alpha Karpenter-for-CAPI, and vCluster's Auto Nodes each actually deliver today.