Tag: homelab

Ceph for a 3-Node Home Lab: When It Actually Pays Off

Ceph on 3 nodes: real hardware requirements, honest performance tradeoffs, and exactly when distributed storage beats a plain NFS share for your home lab.

Kdenlive Proxy Editing on Weak Hardware

Edit 4K (or chunky 1080p60) on a mini PC without the timeline turning into a slideshow. Kdenlive proxies plus VAAPI/NVENC for cheap homelab editing.

Cosmos vs CasaOS vs Umbrel: All-in-One Home Server OSes Compared

Cosmos, CasaOS, and Umbrel each try to make self-hosting one-click. Here's how they actually compare on UI, security, app catalogs, and the all-important escape hatch when you outgrow them.

Snapper for Btrfs Snapshots on Root Filesystems

Set up Snapper on Btrfs to auto-snapshot before every system update, roll back broken machines in minutes, and never lose a working root filesystem again.

Plex Pass Hits $749. Time for Jellyfin.

Plex tripled the lifetime price to $749. Here's a decision tree plus a working Jellyfin migration plan: Docker, transcoding, clients, watch history.

Kopia Repository Server: Multi-Host Backups Done Right

Stop letting every machine write directly to S3. Kopia repository server gives you shared dedup, per-host auth, and sane maintenance, all in one place.

ZFS Send/Receive Over WireGuard for Off-Site Replication

Pipe ZFS incremental snapshots through WireGuard to a friend's NAS or a remote VPS. Encrypted in transit and at rest, no rsync.net bill or vendor lock-in.

mergerfs + SnapRAID: The Poor Man's Unraid

Skip the paid Unraid license. mergerfs + SnapRAID give you flexible JBOD pooling with parity protection on mismatched drives you already own, for free.

Object Storage on a Pi: SeaweedFS Cluster Walkthrough

Run a real S3-compatible object storage cluster on Raspberry Pi 4s with SeaweedFS, low RAM overhead, fast filer, no Ceph drama or surprise cloud bills.

Bind Mounts vs NFS for Container Storage

Bind mounts are fast and simple; NFS is shared and flexible for containers. Pick wrong and your database corrupts at 2 AM. Here is how to choose wisely.

GPU Passthrough on Proxmox: Run LLMs in a VM

GPU passthrough on Proxmox is the best way to isolate LLM workloads, but it's a minefield of IOMMU groups, vfio-pci binding, and Code 43 errors. This post walks through the whole thing end-to-end.

Hoist: Label-Driven Docker Updates

Per-container control over Docker image updates with labels. Auto-update or notify via Discord, Slack, ntfy, no sidecar needed.