I recently put together a small Ceph lab on Ubuntu using cephadm, mainly to understand how everything fits together in a real environment.
Most guides out there are either too abstract or assume multi-node production setups, so I decided to document a single-node deployment end-to-end — focusing on what actually matters when you're getting started.
What this includes
- Ceph installation (cephadm)
- OSD provisioning from previously used disks
- Network bonding (802.3ad / LACP)
- iSCSI gateway setup
- LVM snapshots for rollback scenarios
Everything is based on Ubuntu Server 22.04 and Ceph Squid.
Repo
Full step-by-step docs here:
👉 https://github.com/EzequielPA4/ceph-infrastructure-docs
Why this might help
If you're:
- testing Ceph in a lab
- learning how OSDs really work
- trying to avoid common cephadm issues
- dealing with reused disks (LVM leftovers, etc.)
this will probably save you time.
Notes from the lab
A few things that are worth calling out:
- Ceph is very strict with disks → anything with existing LVM or FS will be rejected
- Cleaning disks properly is key (wipefs + sgdisk + dd in some cases)
- Bonding (LACP) needs correct switch config or it just won’t behave
- Snapshots with LVM are useful, but you need to monitor usage (
lvs -o +data_percent)
Limitations
This is a single-node setup, so:
- no HA
- no quorum discussion
- focused on learning / testing
If you're running something similar
I'm curious how others are setting this up in lab environments.
- Are you using cephadm or something else?
- Any gotchas with iSCSI gateways?
Feel free to share your setup or improvements.
Top comments (0)