Orchestration — Magnum

Magnum is the OpenStack Container Orchestration Engine (COE) service. It provisions and manages container clusters — primarily Kubernetes — on top of OpenStack infrastructure. Magnum uses Heat to create the underlying Nova instances, Neutron networks, Cinder volumes, and load balancers, then installs and configures the chosen COE on top of them. The result is a fully functional Kubernetes cluster whose kubeconfig is returned to the user.

Magnum has been a core OpenStack service since the Liberty release (2015). Kubernetes is the primary supported COE; Docker Swarm and Mesos are deprecated and removed in 2024.x+ releases.

When to Read This Skill

Provisioning Kubernetes clusters on OpenStack with a single CLI command
Creating and managing cluster templates (standard, reusable cluster configurations)
Scaling cluster node groups up or down
Upgrading clusters to a new Kubernetes version
Retrieving kubeconfig credentials for a cluster
Understanding how Magnum generates Heat stacks for cluster provisioning
Configuring cluster auto-scaling and auto-healing
Managing node groups (multi-node-group clusters with mixed flavors)
Configuring magnum.conf for production deployments
Troubleshooting cluster creation failures via Heat stack events

Sub-Files

File	What It Covers
architecture.md	magnum-api, magnum-conductor; cluster templates; clusters and node groups; COE support (Kubernetes focus); Heat stack integration; cluster drivers
operations.md	CLI: cluster templates, clusters, node groups, resize, upgrade, kubeconfig; config file reference
internals.md	Cluster driver interface; Heat template generation; node group lifecycle; cluster state machine; certificate management; auto-healing; auto-scaling

Quick Reference

# Create a cluster template
openstack coe cluster template create k8s-template \
  --image fedora-coreos-40 \
  --keypair mykey \
  --external-network public \
  --dns-nameserver 8.8.8.8 \
  --master-flavor m1.large \
  --flavor m1.medium \
  --network-driver flannel \
  --coe kubernetes

# Create a Kubernetes cluster
openstack coe cluster create my-k8s \
  --cluster-template k8s-template \
  --master-count 3 \
  --node-count 5

# Watch cluster status
openstack coe cluster show my-k8s

# Get kubeconfig
openstack coe cluster config my-k8s
export KUBECONFIG=$(pwd)/config
kubectl get nodes

# Resize worker nodes
openstack coe cluster resize my-k8s 10

# List node groups
openstack coe nodegroup list my-k8s

# Add a GPU node group
openstack coe nodegroup create my-k8s gpu-workers \
  --node-count 2 \
  --flavor m1.xlarge-gpu \
  --image fedora-coreos-40

# Delete a cluster
openstack coe cluster delete my-k8s

Dependencies

Dependency	Purpose	Notes
Keystone	Authentication, service catalog, trust delegation	Magnum creates Keystone trusts for cluster operations
Heat	Infrastructure provisioning via generated HOT templates	All cluster infrastructure is a Heat stack
Nova	Kubernetes master and worker node VMs	Heat provisions Nova servers for all nodes
Neutron	Cluster networking (tenant net, LB, floating IPs)	Per-cluster network with optional Octavia LB for the API server
Cinder	etcd volumes and persistent volume claims (via CSI)	Optional; recommended for etcd durability
Octavia	Load balancer for the Kubernetes API server	Optional; used when `master-lb-enabled=true`
Barbican	Cluster TLS certificate storage	Optional; used when `cert-manager-api=true`
Glance	Cluster node OS images (Fedora CoreOS, Ubuntu, etc.)	Images must have Kubernetes pre-installed or use cloud-init
RabbitMQ (or oslo.messaging backend)	RPC between magnum-api and magnum-conductor	Required
MariaDB 10.6+ or PostgreSQL 14+	Magnum database (cluster templates, clusters, certificates)	Required
`python-magnumclient` / `python-openstackclient`	CLI	`openstack coe *` commands

Releases Covered

This skill covers OpenStack 2025.1 (Epoxy), 2025.2, and 2026.1.

Magnum release notes: https://docs.openstack.org/releasenotes/magnum/