- Shipped
- August 3, 2026 at 1:34 AM UTC
- Author
- Kamo
- Commit
- a1139a0
deploy_k1m1 had never succeeded (0/5 since the repo was created). The job rewrote the kubeconfig API server to https://k1m1:6443, but the runner's job container resolves through public DNS (1.1.1.1) and cannot resolve a LAN hostname, so every deploy died with: dial tcp: lookup k1m1 on 1.1.1.1:53: no such host Reproduced both ways in a container on the host: k1m1:6443 fails, and 192.168.4.22:6443 returns the deployment. That IP is in the API server cert SANs, so TLS still verifies without --insecure-skip-tls-verify. Sibling repos (ESigService, kamo-marketing) never rewrote the server at all — this repo was the outlier. Also adds a fail-fast connectivity probe so a future regression reports the unreachable server instead of surfacing as a confusing kubectl DNS error.