Learn MicroCloud - Installation & Prerequisites

#Introduction

In episode 2 we understood the architecture. Now it's time to install all the components on every node. In episode 3 we install the four snaps, make sure the services are active, and prepare the node prerequisites: the correct Ubuntu distribution, synchronized time, unique hostnames, and empty disks for Ceph OSDs. This is the gateway to the initialization moment in episode 4.

Why prepare the prerequisites first? Because MicroCloud automates configuration, not guesswork. If two nodes share a hostname, their clocks differ, or the disk you point to isn't empty — the init/join process will fail midway. Five minutes of preparation here saves an hour of troubleshooting in episode 16.

#Installing Snap on All Nodes

#Install the Four Snaps

On every node, run:

sudo snap install lxd microceph microovn microcloud

This command installs all four snaps at once. Because all snaps ship versions released together, you automatically get a tested combination: MicroCloud 3.2 with LXD 6.8, or MicroCloud 2.1.3 LTS with the LTS track — depending on the channel you choose.

sudo snap install lxd --channel=6.0/candidate
sudo snap install microceph microovn microcloud --channel=2/stable

#Make Sure Snap Services Are Active

After installation, verify the services are running:

snap services lxd microceph microovn microcloud

The output will show active for lxd, microceph, and microovn. The microcloud snap is on-demand: it doesn't run continuously, but appears when you run a microcloud ... command. This is normal and intentional — the orchestrator is only needed during configuration.

snap list lxd microceph microovn microcloud

#Node Prerequisites

#Ubuntu 22.04+ / 24.04 LTS

MicroCloud supports Ubuntu 22.04 and 24.04 LTS. For this series we use 24.04 LTS as our reference:

lsb_release -a

If a version is already installed, make sure packages are up to date before continuing:

sudo apt update && sudo apt upgrade -y

#Time Synchronization: Chrony

Ceph and LXD clusters are very sensitive to time differences between nodes — clock skew can cause an OSD to be ejected from the cluster or a join to fail. Set up synchronization with chrony:

sudo apt install -y chrony
sudo systemctl enable --now chrony

Verify synchronization:

timedatectl status
chronyc tracking | grep -E "Leap|Stratum"

All nodes should point to the same time source (the same NTP pool or an internal time server). The offset should ideally be under 100 ms — a prerequisite we check again in episode 16.

#Unique Hostnames

Each node needs a unique hostname within a single cluster. Duplicate hostnames make cluster members mix each other up:

sudo hostnamectl set-hostname node-a    # node-a
sudo hostnamectl set-hostname node-b    # node-b
sudo hostnamectl set-hostname node-c    # node-c

Make sure the hostname is truly registered, and ideally resolvable between nodes:

hostnamectl hostname
getent hosts node-a

If there's no DNS, add entries to /etc/hosts on every node so the node names resolve.

#Empty Disk for the Ceph OSD

MicroCeph needs at least one empty disk per node to turn into an OSD (Object Storage Daemon). This disk will be formatted — so make sure it's truly empty:

sudo lsblk -o NAME,SIZE,TYPE,FSTYPE,MOUNTPOINT

Look for a disk without an FSTYPE and without a MOUNTPOINT — for example /dev/sdb or /dev/nvme0n1. Note its path; we'll use it in episodes 4 and 6.

#Multi-node vs Single-node Preparation

• 3 nodes (production): run all the steps above on all three nodes, including installing the same snaps.
• 1 node (testing): install the snaps on a single machine only, but still prepare an additional disk and follow episode 4 in single-node mode.

hostnamectl hostname && lsb_release -d
snap list lxd microceph microovn microcloud
snap services lxd microceph microovn
chronyc tracking | grep -E "Leap|Stratum"
sudo lsblk -o NAME,SIZE,TYPE,FSTYPE,MOUNTPOINT

#Common Pitfalls

• Snap not active after installation: wait a few seconds and run snap services again; some services need startup time.
• Inconsistent channels between nodes: make sure all nodes use the same channel, otherwise versions will differ and the join can fail.
• Full disk in /var/snap: snap stores images and data in /var/snap; make sure there's enough space, especially for the Ceph OSD wal/db (episode 18).
• Forgot chrony: install and enable it before episode 4; a Ceph cluster without time synchronization is prone to latent issues.

#Closing

Key takeaways:

• Install lxd, microceph, microovn, and microcloud via snap on all nodes.
• snap services should show lxd, microceph, and microovn as active.
• Set up chrony for time synchronization; the offset between nodes must be small.
• Unique hostname per node; name resolution between nodes is guaranteed (DNS or /etc/hosts).
• Provide an empty disk per node for the Ceph OSD — never point at a disk containing data.

In the next episode, we'll do the initialization: microcloud init — building the cluster from the first node, choosing storage (disk for Ceph vs local dir), choosing the network (bridge/OVN), waiting for other nodes to join, and verifying the results with microcloud status. This is the first moment your cloud comes alive!