Case studies
Incidents from my homelab, written up in post-mortem format: symptoms, investigation (including the wrong turns), root cause, fix, and what I'd do differently. Some of this work was AI-assisted; the diagnosis and decisions described are mine, and I can defend every one of them.
-
A VPN sidecar that cut two nodes off the internet
How a container's kill-switch firewall escaped its pod and locked down the physical node — twice in one evening — and what that taught me about blast radius.
-
The day my control plane stopped accepting writes
Reads worked, writes hung. Tracing k3s write timeouts to workloads I'd let land on the control plane — and recovering from the reboot that followed.
-
A game server that answered every ping — with a broken reply
Proving a fault was an upstream server regression, not my network, by capturing packets at the right layer — then working around it with a protocol-level fix in a UDP relay.