Infrastructure
Kubernetes, networking, DNS, containers and the machines agents run on. 10 posts, newest first.
Posts
Removed ranges kept on the Service: loadBalancerSourceRanges in Kamaji
Clients outside a TenantControlPlane's old source ranges stayed blocked after the ranges were removed from it.
An update checked against the old object: the DNS service IP webhook in Kamaji
A TenantControlPlane could be edited to a DNS service IP outside its service CIDRs.
An error that was logged, not returned: the DataStore lookup in Kamaji
Kamaji hosts Kubernetes control planes as pods.
A log line that read the wrong error: the API server certificate check in Kamaji
A one-line mistake in a log call made Kamaji's reconcile panic when apiserver.crt was missing from the API server certificate Secret.
A reset that skipped a removed block: kube-apiserver extraArgs in Kamaji
Deleting the whole extraArgs block from a TenantControlPlane in Kamaji, which runs Kubernetes control planes as pods, left its flags on the kube-apiserver container.
Two goroutines, one receive: a leak in the k3s tunnel proxy
k3s leaked a goroutine, its 32 KiB buffer and both connections whenever the two copy directions of a tunnel connection failed at about the same time.
Encrypting every record to read one name: a fix in external-dns
external-dns synchronizes Kubernetes Services and Ingresses with DNS providers.
A verify target that rewrote the tree: four review points on a cri-tools script change
make verify-docs and make verify-go-modules in cri-tools rewrote files in your working tree, then failed if git status showed anything dirty.
One String() call per pod: 320x faster in one kube-router function
I got a small change merged into kube-router (PR #2192). The code change deletes five lines; I also added a test and a benchmark.
Agents on shared machines: four ways they quietly hurt the host
I run a fleet of coding agents on machines I own. Most are shared: agents, my own work, and a few long-running services on the same box.