Deploy Proxmox Infra / Pulumi Preview (pull_request) Successful in 53s
Deploy Proxmox Infra / Pulumi Deploy (pull_request) Skipped
Deploy k8s Bootstrap / Pulumi Preview (pull_request) Successful in 40s
Deploy k8s Bootstrap / Bootstrap k3s Cluster (pull_request) Skipped
Provider, image download, and template were three separate consts per physical node with matching lookups scattered through nodeConfigs - easy to update one and miss another when adding a node. Folded them into one pveNodes pipeline (seed data -> provider -> image -> template) and had nodeConfigs reference a host row instead of copying its fields. Also folds in two fixes discovered while bringing up worker-3, now inseparable from the restructuring above: - disable the default physical CD-ROM passthrough (cdrom: fileId "none") that failed VM start on nodes without an optical drive - shorten the qemu-agent timeout from the 15m default so refresh/read fails fast instead of hanging