Learn MicroCloud - Multi-cluster: MicroCloud Cluster Manager

#Introduction

So far you've managed a single MicroCloud. But real edge organizations often have many clusters — one per branch, one per site, one per environment. Managing them all from different terminals is a recipe for chaos. In episode 12 we cover MicroCloud Cluster Manager, a tool that unifies many MicroClouds into a single dashboard.

An analogy: MicroCloud Cluster Manager is a regional headquarters for many branches (clusters). Each branch still handles its own daily operations, but reports, health, and alarms are centralized in one place. You see all stores from one screen, know which one is having problems, and can dispatch help without calling each store one by one.

#What Is MicroCloud Cluster Manager

#Status and License

MicroCloud Cluster Manager is a tool that manages many MicroCloud clusters from a single interface:

• Status: still beta — not for critical production (as of March 2026, integrated with MicroCloud 3.1+).
• License: AGPLv3 — open source, with a strong copyleft requirement.
• Language: written in Go.
• Stack: built on Juju (Charmed Operators) with PostgreSQL as the central database and Traefik as the ingress.

           ┌───────────────────────────┐
           │  Cluster Manager UI/API   │
           │  (Go · AGPLv3 · Juju)     │
           │  PostgreSQL · Traefik     │
           └────────┬─────────┬────────┘
                    │         │
           ┌────────▼───┐  ┌──▼─────────┐
           │ MicroCloud │  │ MicroCloud │
           │  cluster 1 │  │  cluster 2 │
           └────────────┘  └────────────┘

#Why Juju?

The choice of Juju isn't accidental — Canonical has long used the Charmed Operators model to operate complex services (OpenStack, Kubernetes). Cluster Manager inherits the same model: every service (PostgreSQL, Traefik, controller) runs and is maintained as a charm, so upgrades and recovery can be automated in a consistent way.

#Managing Many Clusters from One UI

#Cluster Registration

Once Cluster Manager is active, you register each MicroCloud you want to manage. Cluster Manager connects to the LXD API of each cluster (HTTPS, episode 14) and collects its status:

• Cluster members and their roles.
• Ceph OSD status and storage usage.
• OVN network status.
• Instance health.

From one UI, you see the entire fleet: which site is HEALTH_OK, which is degraded, which needs attention.

#Centralized Operations

Although still beta, Cluster Manager is already heading toward centralized operations across clusters — monitoring and, as its roadmap matures, operating resources from one place. For now, observability is its main value: a quick answer to "what's happening across all sites?".

#Cross-cluster Observability

#One View, Many Sites

With many clusters, the common problem is scattered data — each cluster has its own logs and status. Cluster Manager gathers it all into one layer:

• Health summary per cluster.
• Alarms for critical conditions (node down, OSD down, degraded cluster).
• Drill-down from a concise status to per-node detail.

#The LXD Grafana Dashboard

Cluster Manager also integrates with the LXD Grafana dashboard — the official dashboard that visualizes LXD cluster metrics (instances, storage, network). We explore the dashboard and full monitoring in episode 20.

MicroCloud cluster ──► exporter (Prometheus) ──► Grafana dashboard
        └──────────► Cluster Manager UI/API ──► status fleet

#When to Use Cluster Manager

• Many edge clusters: 5+ sites each running their own MicroCloud.
• Centralized ops teams: admins monitoring a fleet from one location.
• Cross-site observability: a need for centralized health reporting.
• Not yet suitable for: production requiring full stability, or small fleets (1-2 clusters) that can be monitored fine with microcloud status.

#Common Pitfalls

• Relying on beta in production: don't — use standard exporters until Cluster Manager reaches a stable release.
• Version mismatch: Cluster Manager requires MicroCloud 3.1+; 2.x LTS clusters aren't fully integrated yet — check the release notes.
• Forgetting the basic observability layer: a UI doesn't replace Prometheus exporters and alerting — both run side by side.
• One password for everything: Cluster Manager gives a view into the whole fleet — secure its UI access like admin access (episodes 14-15).

#Closing

Key takeaways:

• Cluster Manager (beta, AGPLv3, Go) manages many MicroClouds from one UI.
• Stack: Juju + PostgreSQL + Traefik; integrated since MicroCloud 3.1+.
• Its main value today: cross-cluster observability and the LXD Grafana dashboard.
• Don't rely on beta in production — use standard Prometheus exporters.
• Secure access to Cluster Manager because it sees the entire fleet.

In the next episode, we'll cover snap security & confinement — how every component runs under strict confinement, automatic transactional updates with rollback, and best practices: not installing components outside of snaps and reviewing snap interfaces. Security starts taking the stage!