12 min read
Reach services across clouds with internal load balancers
One load balancer per service, or one Traefik routing by hostname. What each costs, why GKE pods need a masquerade fix, and how names resolve.
Infrastructure
11 posts tagged “Infrastructure”.
10 min read
Connect GKE to AKS over a VPN, and prove it works
Build the tunnel between Google Cloud and Azure with BGP, advertise only what should cross, then test four layers until real traffic flows.
23 min read
Building an AI hyperscaler on Kubernetes
How an AI cloud fits together on Kubernetes, from what customers actually buy and the seven layers between datacenter and console to four products on one GPU pool and how tenants share it safely, the lessons AWS and Google learned about cells and control planes, and what changes with scale.
8 min read
Moving Kubernetes workloads across an air gap with Hauler
Declare images and charts in one manifest, package them into a single archive, carry it across, and serve it from a binary with no dependencies.
8 min read
Scanning images, filesystems and IaC with Trivy
One scanner for container images, dependencies, Terraform and secrets: install it, read its output, and fail a build on the right things.
13 min read
Building Kubernetes-as-a-Service from bare metal
What it takes to run managed Kubernetes when you are the cloud, from hosted control planes, cells and etcd to networking, storage, upgrades and tenant isolation without cloud APIs, plus a version small enough to build on a laptop.
4 min read
Upgrading a cluster and backing up etcd
Upgrade a kubeadm cluster one minor version at a time, control plane first, then take and restore etcd snapshots, because etcd holds the cluster's entire state.
5 min read
Building a cluster with kubeadm
Prepare the nodes, install a container runtime and the Kubernetes packages, bootstrap the control plane, add a network plugin and join the workers. Then what each core component does.
4 min read
Persistent storage: volumes, PVs, PVCs and StorageClasses
A container loses its files when it stops. Volumes, PersistentVolumes, PersistentVolumeClaims and StorageClasses are how Kubernetes keeps data that has to outlive it.
1 min read
The part of a migration nobody schedules
The migration itself takes a week. Deleting the old thing takes a year.
6 min read
Create a pre-signed URL for a Civo object store
Upload a file to a Civo object store with s3cmd, then hand out a link that works for exactly one hour and then stops.