bootstrap.sh no longer needs the provisioning network: it installs and
starts the region on the management address; --iface later models the
subnet, serves DHCP and moves server_url. rollcall net discover shows what
each interface sees (LLDP, DHCP servers via a parallel probe, neighbours)
and suggests the next step. commission_append follows server_url. Doctor
treats a host that serves no DHCP yet as day 0. Fixes found on a clean VM:
os-release clobbered the rack name, install.sh not executable, slow DHCP
offers missed.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
bootstrap.sh sets up a region on a fresh Debian 13 host (packages,
PostgreSQL, TFTP tree, config, services, verified commissioning image,
provisioning subnet with ranges/DHCP, optional NAT, first OS image).
rollcall doctor checks an installation and says what to fix, including a
DHCPDISCOVER probe for foreign DHCP servers. pxelinux-default no longer
hard-codes the lab address.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- fabrics, VLANs, subnets, IP ranges (dynamic/reserved/static), IPAM with adoption of
addresses machines already have; rollcall fabric|vlan|subnet|range|ip
- interface layouts (physical/bond/vlan, NICs by selector: boot, name, mac, fastest:N, ib:N)
per machine or per tag; deployment allocates addresses and writes cloud-init network-config
plus udev name rules (RHEL cloud images boot with net.ifnames=0); release frees them;
rollcall layout, rollcall net show|set
- region/rack: rollcall rack long-polls its desired state (per-MAC PXE files kept in the
database, dnsmasq configuration and host files) and a BMC command queue, caches boot and OS
images (kernels/initrds prefetched for TFTP), proxies the node API and tags requests with its
name; the region runs an embedded rack controller; install-rack.sh, rollcall-rack.service
- dnsmasq configuration generated per rack from the VLANs it serves (split pool for two racks);
reload script restarts only on change and restores the last working config if dnsmasq does
not come back (drilled in the lab)
- web: Network page, per-machine network tab, layout action, tag layout; console: Network page,
network section in the machine view, layout action
- tests for candidates, plans (default, bond+VLAN, pinned, DHCP fallback), network-config,
udev rules and dnsmasq rendering
- lab: pve-vbmc address script finds the interface per BMC address
Verified in the lab: embedded rack (10.213.77.0/24) and a remote rack controller on a second
SDN segment (10.213.78.0/24): enlistment, commissioning, BMC through the rack's queue, image
cache, deployment and phone-home through the rack; Debian 13, Ubuntu 24.04 and Rocky 9 with a
static boot NIC and with an active-backup bond carrying VLAN 200.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Profiles bind a RAM-booted root image to overlay templates; assigning a
machine allocates a node name and management IP, writes the per-MAC PXE
entry with a per-node token and exports static leases for dnsmasq. Nodes
pull rendered overlays (hosts, slurm.conf from inventory, gres.conf, SSH
host key, munge key) through rollcall-overlay.service; the runtime refresh
timer is the heartbeat. Adds the compute-debian13 image definition and a
generic live-build root image script, CLI and web UI for profiles/assign.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Go server (API, embedded web UI, CLI) backed by PostgreSQL with AES-GCM
sealed BMC secrets; Debian live commissioning agent and live-build recipe;
Proxmox lab with pve-vbmc virtual BMCs; Gitea workflows for CI, releases
and deployment; project website.
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
ilkerandClaude Fable 5.1