What it is
Proxmox VE is a Debian-based type-1 hypervisor with a web UI on top. I run it on two physical machines clustered together — an old Optiplex and a beat-up laptop — and manage everything from a single browser tab. It hosts a mix of LXC containers (for everything that doesn't need its own kernel) and full VMs (for the few things that do).
Why I run it
I wanted something more serious than "Docker on a VPS" without paying for VMware. The killer feature is the unified web UI: one URL gives me console access, snapshots, backup scheduling, live migration, and resource graphs for every guest across both nodes. LXCs are the daily workhorse — they boot in under a second and feel like processes, not VMs — and when I need a real kernel (Home Assistant OS, for example), a full VM is one click away in the same UI.
How I use it
Every service in this catalog lives on one of the two nodes. The split is roughly:
- Optiplex — primary workloads I want stable and rarely rebooted. Vaultwarden, Home Assistant, the Docker edge container that holds my reverse proxy, the Tailscale gateway.
- Laptop — anything I/O-heavy or storage-heavy, because that node has a 1 TB NVMe data drive. Immich, Nextcloud, n8n, Proxmox Backup Server, plus media-tools and dev-tools Docker LXCs.
The cluster is small enough that I don't bother with HA failover; the value of clustering for me is the single pane of glass, not auto-recovery.
Setup notes
- Hardware: Optiplex i5-8500 / 24 GB / single SSD. Laptop i7-8750H / 32 GB / NVMe boot + 1 TB WD_BLACK SN850X data drive.
- Storage philosophy: backups never live on the same physical disk as the data they protect. An external NVMe enclosure (always plugged in) is the PBS datastore — separate disk, physically grabbable in an emergency.
- Updates: manual, monthly. I treat hypervisor upgrades as planned events, not background activity.
Runbook
- Healthy looks like: both nodes green under Datacenter → Cluster, all guests showing running, no red alerts on the node summary.
- Node crash recovery: wait two minutes for boot and cluster sync, then audit each guest. Check
journalctl -b -1on the recovered node for the root cause — kernel panic, OOM-killer, thermal event, disk error. - Common gotcha: by default, Proxmox does not auto-start guests at boot. Each LXC and VM has its own "Start at boot" toggle under Options. After my first node crash I found half my services stayed off until I checked them one by one. Audit the flag for anything critical, and use the Start/Shutdown order field if guests need to come up in sequence (network gateway before the things that depend on it).
- Where logs live:
journalctlon the node for system stuff,pct enter <ID>to drop into an LXC's shell, the per-guest Console tab in the web UI when the network is the problem.