This draft captures the practical work of creating a new cloudflared tunnel, replacing the Kubernetes secrets that hold the certificate and tunnel credentials, and updating the config map and deployment to run the new tunnel.
This draft turns a noisy environment-specific note into a cleaner pattern: create the Cloudflare Tunnel credentials, store them as Kubernetes secrets, define a cloudflared config map, run the deployment, and validate that the tunnel actually maps the hostname to the intended TCP service.
This draft turns a real internal outage note into a cleaner incident write-up covering the forced Kubernetes upgrade, the storage failure that followed, the backup gap we ran into, and the operational changes that came out of it.
This draft captures the part of a Longhorn installation that usually matters most in practice: make sure the node dependencies are in place, enable iscsid, run the environment check script, fix the obvious blockers, and only then install the chart.
This draft captures the practical shape of a cluster-side Cloudflare tunnel setup: install cloudflared, register the service, and use DNS-backed certificate workflows where needed.
This draft turns a real cluster note into a cleaner walkthrough for installing cert-manager, configuring a Cloudflare-backed ClusterIssuer, issuing certificates, and handling a few common ingress-related surprises.
This draft turns a real outage recovery note into a readable operations walkthrough covering service scale-down, Elasticsearch cleanup, database restore, and controlled service startup.
This draft is based on a real cluster note where I validated private image pulls in Kubernetes by creating a Docker registry secret, attaching it to a deployment, and testing the workload in a dedicated namespace.
This draft captures a Kubernetes preparation issue where sysctl reloads failed because /proc/sys/net/bridge/bridge-nf-call-iptables did not exist until br_netfilter was loaded.
This draft captures a real cluster recovery workflow where problem nodes were cleaned up, re-registered into Rancher, and then brought back into a larger rebuild that included storage, MetalLB, and GPU-related follow-up work.