Learn MicroCloud - Core Concepts & Main Architecture

#Introduction

In episode 1 we understood why MicroCloud exists. Now we dissect how it works: what its components are, how they connect, and who does what. In episode 2 we build a mental architecture map — the foundation that makes every command in the following episodes feel logical, not merely memorized.

A simple analogy: MicroCloud is a project team with three specialist members. LXD is the contractor managing the "buildings" (containers and VMs). MicroCeph is a distributed warehouse that stores all materials with backup copies. MicroOVN is the road network connecting all the buildings. And the microcloud snap is the project manager coordinating them to work as one unit.

#Three-Layer Architecture

#LXD: Virtualization and Compute

LXD is an image-based hypervisor that manages two kinds of instances:

• System containers: containers that behave like virtual machines — with their own systemd, init, and services — but share the host kernel. Lightweight and fast to boot.
• Virtual machines (VMs): instances with their own guest kernel using QEMU/KVM, for workloads that need full kernel isolation or live migration.

LXD provides a REST API (the lxc client) that serves as the admin's main interface: lxc launch, lxc list, lxc snapshot, up to lxc cluster list. Inside a MicroCloud cluster, LXD is also responsible for instance scheduling across nodes and cluster leader election.

#MicroCeph: Distributed Storage

MicroCeph wraps Ceph — distributed storage proven over decades in OpenStack and large clouds — into an easy-to-operate snap. Ceph provides three interfaces from one pool of hardware:

• RBD (block): raw disks for LXD instances — the most commonly used.
• CephFS (file): a distributed filesystem for sharing files.
• RGW (S3 object): S3-compatible object storage for modern workloads.

Data is replicated across nodes (default replication factor 3 on a 3-node cluster), so if one node dies, the data stays safe. We dive deeper into MicroCeph in episode 6.

#MicroOVN: SDN and Networking

MicroOVN wraps OVN/OVS (Open Virtual Network / Open vSwitch) into a snap. OVN is an SDN that provides:

• Logical switches: virtual L2 networks — each LXD network gets one.
• Logical routers: routing between virtual networks.
• Overlay tunnels: inter-node traffic is wrapped in tunnels (VXLAN/Geneve) so instances can "move" between nodes without changing their IP addresses.

MicroOVN automatically registers networks with LXD, so you can use networks from instances right away. We dissect it in episode 7.

#MicroCloud: The Orchestrator

The microcloud snap is the conductor that doesn't provide storage or networking itself — it orchestrates the three snaps above. Its main tasks:

• Running microcloud init/join to build the cluster.
• Discovering other nodes on the network (via mDNS).
• Configuring LXD, Ceph, and OVN so they're consistent with each other.
• Providing microcloud status and microcloud cluster for operations.

┌─────────────────────────────────────────┐
│                snap microcloud          │  orkestrator
├──────────────┬──────────────┬───────────┤
│  snap lxd    │ snap microceph│ snap microovn│
│  compute     │ storage      │ SDN       │
├──────────────┴──────────────┴───────────┤
│            kernel + hardware            │
│   CPU/RAM  ·  disk OSD  ·  NIC/bridge   │
└─────────────────────────────────────────┘

#Components and CLI

#Installed Snaps

Every node in the cluster has these four snaps installed and running:

snap list
snap services lxd microceph microovn microcloud

snap services should show the active status for lxd, microceph, and microovn (the microcloud snap runs on-demand when invoked).

#The microcloud CLI

The main microcloud commands and their functions:

• microcloud init — builds a new cluster from the first node.
• microcloud join — adds additional nodes to the cluster.
• microcloud status — cluster status summary: members, storage, networking.
• microcloud cluster — cluster management subcommands (list members, etc.).

microcloud --help
microcloud status

#Supporting CLIs

Beyond microcloud, you'll use three other CLIs regularly:

• lxc — manages instances, networks, storage, profiles, and the LXD cluster.
• microceph — adds disks (OSDs), views the Ceph cluster status.
• microovn — views OVN status and configuration.

microceph --help
microovn --help
lxc --help

#The LXD API

All LXD interactions go through the REST API, exposed via a unix socket (/var/snap/lxd/common/lxd/unix.socket) or HTTPS (episode 14). The lxc CLI is just a client of this API. That means you can automate MicroCloud with any script or application that speaks HTTP — for example, a Terraform provider (episode 19).

#Data Flow: One Request, Three Layers

To understand how all three come together, follow the flow when you run lxc launch ubuntu:24.04 c1:

1. LXD receives the request, picks a node based on scheduling, and fetches the image from the image store.
2. MicroCeph provides a block volume from the RBD pool for the instance's root disk — replicated to other nodes.
3. MicroOVN provides the logical network — the instance gets an IP from the network already auto-registered with LXD.
4. LXD creates the container/VM on the chosen node and connects it to that storage and network.

The result: one command, and the three layers work together transparently behind the scenes.

#Conceptual Pitfalls

• Thinking MicroCloud is a hypervisor: MicroCloud isn't a hypervisor — LXD (and QEMU underneath it for VMs) provides the compute. MicroCloud is the orchestration layer.
• Expecting a separate "controller": MicroCloud has no central controller; all nodes are equal and the cluster maintains consistency via quorum.
• Treating Ceph as just RAID: Ceph replicates data across nodes (not within a single machine), so node memory and network directly affect storage performance.

#Closing

Key takeaways:

• LXD = compute (system containers + VMs), MicroCeph = storage (block/file/object), MicroOVN = SDN.
• The microcloud snap orchestrates all three without being a hypervisor or controller of its own.
• Main CLIs: microcloud (init, join, status, cluster), lxc, microceph, microovn.
• The LXD API is the universal interface — lxc is just one of its clients.
• A single lxc launch request involves LXD (scheduling), Ceph (volumes), and OVN (networking) at once.

In the next episode, we start taking action: installation & prerequisites — installing lxd, microceph, microovn, and microcloud via snap on all nodes, making sure snap services is active, setting up synchronized time with chrony, unique hostnames, and empty disks for Ceph OSDs. Get your nodes ready!