Skip to content

Glossary

Terms, acronyms, and homelab-specific naming that recur across the Infrastructure documents.

Homelab naming

  • mainframe — the primary Proxmox node. Site of most of the Ceph monitor/OSD incidents and the original VM 100 builds.
  • pve1 / pve2 / pve3 / pve4 — the other Proxmox cluster nodes.
  • VM 100 / debian-docker — the main Docker host VM. Rebuilt from scratch more than once; the subject of most Automation and Containers & Apps documents.
  • VM 110 — a second Docker VM provisioned by reusing VM 100's cloud-init pattern.
  • snips — the author's dedicated Proxmox storage ID for cloud-init snippets. Backed at different points by a local directory and by CephFS.
  • cephpool — the Ceph RBD storage pool used for VM disks.

Proxmox & Ceph

  • Proxmox VE — the hypervisor platform used throughout.
  • qm — the Proxmox CLI for managing VM configuration, disks, and lifecycle (qm set, qm config, qm cloudinit update, …).
  • cloud-init — the guest first-boot provisioning system. The single biggest recurring source of trouble in these documents — see Recurring Patterns.
  • cicustom — the qm set field that points a VM at custom cloud-init user-data and network-data snippet files, e.g. user=snips:snippets/docker-userdata.yml.
  • snippets (storage content type) — the Proxmox storage content type for cloud-init custom files. A storage with path /foo and content snippets expects files under /foo/snippets/, referenced as storageID:snippets/<file>.
  • pvesm — the Proxmox storage manager CLI, used constantly to resolve what a storage ID actually maps to on disk (pvesm list, pvesm path).
  • RBD (RADOS Block Device) — Ceph's block storage, used for VM disks and the cloud-init disk itself.
  • CephFS — Ceph's file-based storage, used here for snippets and for a shared QCOW2/ISO image library.
  • OSD / MON / MGR — Ceph's core daemons: Object Storage Daemon (holds data), Monitor (cluster state and quorum), Manager (cluster metrics and orchestration).
  • HA (High Availability) — Proxmox's VM failover mechanism. Restarts a VM on another node after the original node is lost; this is a restart, not a live continuation of the VM's running state — that distinction matters and is easy to get wrong.
  • Live migration — moving a running VM to another node while the source stays healthy. With Ceph-backed storage, only VM memory state transfers; the disk doesn't need to move.
  • QMP (QEMU Machine Protocol) — the control protocol Proxmox uses to manage a running VM; a frequent source of stale-lock and timeout symptoms during VM start/stop under load.
  • EFI / OVMF — UEFI firmware. Some appliance images (Home Assistant OS) require it instead of legacy BIOS, which shows up as a VM stuck at "Booting from hard disk..." until the firmware is switched.

Docker & reverse proxy

  • Traefik — the reverse proxy used across the homelab. Migrated from v2 to v3 more than once, usually forced by a Docker Engine upgrade breaking the Docker-provider API version.
  • Gluetun — the VPN container that download clients (qBittorrent, Deluge, SABnzbd) share a network namespace with via network_mode: service:gluetun.
  • docker-socket-proxy — a sidecar that proxies the Docker socket to Traefik instead of exposing /var/run/docker.sock directly.
  • traefik-proxy — the shared external Docker bridge network that Traefik and proxied application stacks all attach to.
  • Bind mount — a host-path-to-container-path mapping, as opposed to a named Docker volume. Used throughout for appdata and Compose file locations (/opt/docker-apps, /opt/compose).
  • PUID / PGID — the UID/GID environment variables LinuxServer-style containers use to match host file ownership; a recurring source of permission errors on bind-mounted paths.

Networking

  • OPNsense — the firewall/router platform.
  • ONT (Optical Network Terminal) — the ISP hand-off device for fiber; distinguishing an ONT-level outage from an OPNsense misconfiguration is a recurring first triage step.
  • Storm control — switch-level flood limiting (broadcast/multicast/ unknown-unicast) used to stabilize a flat LAN segment.
  • Split DNS — resolving a public hostname to a LAN IP for internal clients, used so Traefik-proxied services keep valid TLS hostnames on and off the LAN.