Kubernetes
Kubernetes the Hard Way, Part 3: Back Up etcd and Restore a Lost Control Plane
Back up a Kubernetes 1.37 cluster by hand with etcd snapshots, then delete the control plane and restore it. Includes the encryption key trap, all tested.
Topic
Engineering
12 articles tagged Engineering, newest first.
Kubernetes
Kubernetes the Hard Way, Part 5: Certificates, Expiry and Renewal Without Downtime
Read every Kubernetes certificate, watch one expire, and renew them all by hand. Most components reload a renewed certificate with no restart at all. Tested.
Kubernetes
Kubernetes the Hard Way, Part 2: A Highly Available Control Plane
Build a Kubernetes 1.37 control plane that survives losing a server. Three control planes, etcd with TLS, a load balancer, and failure drills, all tested.
Kubernetes
Kubernetes the Hard Way, Part 4: Pod Networking by Hand, Without a CNI Plugin
Build Kubernetes pod networking by hand with network namespaces, veth pairs, a bridge, routes and NAT, then call a CNI plugin yourself and trace a Service. Tested.
Kubernetes
Kubernetes the Hard Way in 2026: Kubernetes 1.37 From Scratch
Kubernetes the Hard Way, updated for Kubernetes 1.37 and Debian 13. Every command tested end to end, on local machines and on Hetzner Cloud.
Kubernetes
We Moved Our Production to UpCloud. Here Is What Sovereignty Actually Cost.
Ankra's own platform now runs in Helsinki. The timeline, the bill, the benchmark numbers, and the places UpCloud let us down, from the team that did the move.
Kubernetes
AWS vs GCP vs Hetzner: Six Clouds, One Kubernetes Benchmark
We ran the same k6, iperf3 and fio suite on EKS, GKE and Hetzner and lined the results up against UpCloud, OVH and DigitalOcean. The cheapest node in the test won more than it had any right to.
Kubernetes
The Spec Sheet Is Not a Benchmark: Testing Kubernetes Clusters in 15 Minutes
Two identical cloud VMs measured 2.8x apart on CPU; one failed etcd's fsync bar. A 15-minute in-cluster benchmark suite catches this before production does.
Kubernetes
UpCloud vs DigitalOcean vs OVH Kubernetes Benchmark
UpCloud vs DigitalOcean vs OVH on identical Kubernetes clusters. UpCloud led on storage and network, OVH on latency, DigitalOcean had the slowest tails.
Kubernetes
Postgres on Kubernetes Is No Longer a Dare
The 'never run a database on Kubernetes' rule predates CSI and operators. What production Postgres takes in 2026, and when managed still wins.
Engineering
Ankra is now ankra.ai: Rewritten in Go, Ready for What's Next
A new name, a new engine, and the biggest release we have ever shipped: the Go rewrite, an AI that owns the incident loop, passkey security, European multi-cloud, and Ankra on your phone.
Engineering
GitOps Without ArgoCD: Inside Ankra's Native Deployment Engine
New Ankra clusters no longer run ArgoCD. Here is the engineering story of the native deployment engine: why we built it, how the reconcile loop works, what the migration does to your live resources (nothing), and what we have not finished yet.
Try it yourself
Run it on your own application.
Connect a repository and a cluster. Ankra builds, deploys and operates it with you.
From something you built to something people use.
- Deploy from your repository
- Run on infrastructure you choose
- Operate with your team and agents