Skip to content

๐Ÿงช Quick lab

Stop 1 ยท Discover

Scripts that build a complete, disposable OpenStack (2026.1) on a cloud VM in about an hour, exercises to break and repair it, and one command that deletes it all.

Pick a lab ยท Commands ยท Exercises ยท Delete it all

Note

Status: early, tested on one cloud. Both variants and all eleven exercises were run end to end between 2026-10-23 and 2026-10-25 by OzarLand's AI lab agents (exercises 10 and 11 were edited after their run and not run again), on one OpenStack cloud (Felcloud, the reference cloud below), with OpenStack 2026.1, Kolla-Ansible 22.2.0 and Ubuntu 24.04, in a project reaching the Internet through an HTTPS proxy. Not tested yet: other clouds, a project with direct Internet access, and a run by a person following the guide. Known script gaps are tracked in issue #25. Reports from your cloud are the most useful contribution.

On a prepaid cloud you cannot spend more than your credit; on any other cloud, set a budget alert first.

Pick your lab

Single VM (openstack-cloud) Multinode (openstack-cloud-multinode)
๐ŸŽฏ What you can practise The core services: DHCP, metadata, Nova, Open vSwitch, security groups and floating IPs (exercises 1 to 6) 3 VMs: live migration, a failed compute, host maintenance (exercises 7 to 9). 5 VMs: also a failed controller, Galera and RabbitMQ, VIP failover (exercises 10 and 11)
๐Ÿ–ฅ๏ธ VMs and quota 1 VM: 4 vCPU, 16 GB RAM, 50 GB disk 3 VMs: 12 vCPU, 32 GB RAM. 5 VMs (CONTROLLERS="3"): 20 vCPU, 64 GB RAM
โฑ๏ธ Time About 1 hour (57 min measured, create.sh to destroy.sh) Measured, create.sh to destroy.sh: 65 min with 3 VMs, about 75 min with 5 VMs
๐Ÿ’ถ Cost of one run 0.19 TND (0.05 EUR); 0.20 TND (0.06 EUR) with a floating IP 3 VMs: 0.52 TND (0.15 EUR). 5 VMs: 1.08 TND (0.32 EUR)
๐Ÿ”ฅ Cost of each forgotten day 4.70 TND (1.37 EUR); 5.06 TND (1.49 EUR) with a floating IP 3 VMs: 11.42 TND (3.34 EUR). 5 VMs: 20.83 TND (6.07 EUR)

๐ŸŒฑ New to OpenStack? Start with the single VM. The costs are the measured duration multiplied by the reference cloud's listed prices on 2026-10-23 (not an invoice); each guide has the details. Felcloud, OzarLand's first sponsor, provides the cloud credit used to build and test OzarLand's labs, and has no say over the content; its prepaid model fits a lab with no surprise bill. No provider is required: every variant works on any OpenStack cloud that meets its requirements. Its pricing page says: "Top up your balance and spend only what you loaded." (checked on 2026-10-23).

Caution

On any cloud: on a prepaid one, load only the credit you are ready to spend; on a pay-as-you-go one, you are billed for every hour the VMs, their volumes and floating IPs exist, so set a budget alert in your provider's console before you start. A variant for a personal machine with KVM, with no cloud cost, is wanted.

๐Ÿ”„ How it works

flowchart LR
    cost["๐Ÿ’ถ Your price per hour<br/>in config.env"] --> create["๐Ÿš€ create.sh<br/>shows the cost, asks,<br/>creates the VMs"]
    create --> deploy["โš™๏ธ deploy.sh<br/>installs OpenStack"]
    deploy --> ex["๐Ÿงช Exercises<br/>break and repair"]
    ex --> destroy["๐Ÿ›‘ destroy.sh<br/>deletes everything"]
    destroy --> check["๐Ÿ” Check that<br/>nothing is left"]
    classDef neutral fill:#f3f4f6,stroke:#9ca3af,color:#111827
    classDef go fill:#dbeafe,stroke:#2563eb,color:#1e3a8a
    classDef lab fill:#dcfce7,stroke:#16a34a,color:#14532d
    classDef stop fill:#fee2e2,stroke:#dc2626,color:#7f1d1d
    class cost,check neutral
    class create,deploy go
    class ex lab
    class destroy stop

create.sh shows the estimated cost per hour and asks before creating anything. deploy.sh deploys OpenStack with Kolla-Ansible. destroy.sh deletes what create.sh created and fails loudly if anything is left.

๐Ÿ’ป The three commands

You need an OpenStack cloud account, its clouds.yaml, bash, ssh, python3 and the OpenStack client: the guide lists them. The scripts are in the repository of OzarLand Academy, under docs/lab/variants/:

git clone https://github.com/OzarLand/ozarland-academy.git
cd ozarland-academy/docs/lab/variants/openstack-cloud    # or openstack-cloud-multinode

In the variant's directory:

cp config.example.env config.env && $EDITOR config.env   # your cloud, network and price per hour
./create.sh     # shows what it will create and the cost, asks, creates
./deploy.sh     # about an hour; run it again after a failure, it resumes
./destroy.sh    # when you are done: deletes everything, after asking

Then do your first exercise: DHCP agents down.

Delete everything when you are done

Caution

๐Ÿ›‘ A lab left running keeps costing money on a pay-as-you-go cloud, and keeps using credit on a prepaid one.

  1. Run ./destroy.sh in the variant's directory. After asking, it deletes everything create.sh created (VMs, key, security group, ports, floating IP, root volumes if any, and the multinode variant's lab network and subnet) and fails loudly if anything is left.
  2. Check in your provider's console, or with openstack server list, openstack volume list and openstack floating ip list, that nothing from the lab is left.
  3. On a pay-as-you-go cloud, check the next day that the usage really stopped.

Every exercise ends with the same reminder: Done with the lab for today? Delete everything: a running lab keeps costing.

๐Ÿงช Exercises

Each exercise breaks one thing on purpose and repairs it with a runbook from openstack-vm-runbooks, or, for 10 and 11, a page of kolla-ansible-deployment-patterns. All the exercises.

flowchart LR
    subgraph one ["๐Ÿ–ฅ๏ธ Single VM"]
        E1["1 ยท DHCP agents down"] --> E2["2 ยท Metadata agents down"] --> E3["3 ยท nova-compute down"]
        E3 --> E4["4 ยท Restart nova_libvirt"] --> E5["5 ยท Restart OVS agent"] --> E6["6 ยท Unreachable from outside"]
    end
    subgraph three ["๐Ÿ–ฅ๏ธ๐Ÿ–ฅ๏ธ๐Ÿ–ฅ๏ธ Multinode, 3 VMs"]
        E7["7 ยท Fence before evacuate"] --> E8["8 ยท Drain for maintenance"] --> E9["9 ยท Live migration"]
    end
    subgraph five ["๐Ÿ–ฅ๏ธร—5 Multinode, 3 controllers"]
        E10["10 ยท Lose one controller"] --> E11["11 ยท Lose two: quorum"]
    end
    one --> three --> five
    classDef s fill:#dcfce7,stroke:#16a34a,color:#14532d
    classDef m fill:#dbeafe,stroke:#2563eb,color:#1e3a8a
    classDef h fill:#fef3c7,stroke:#d97706,color:#78350f
    class E1,E2,E3,E4,E5,E6 s
    class E7,E8,E9 m
    class E10,E11 h
# Exercise Needs Runbook
1 DHCP agents down: a new instance gets no address ๐Ÿ–ฅ๏ธ Single VM New instance gets no IP
2 Metadata agents down: a new instance gets no SSH key ๐Ÿ–ฅ๏ธ Single VM New instance boots without metadata
3 nova-compute down: instances keep running ๐Ÿ–ฅ๏ธ Single VM nova-compute down, instances still running
4 Restarting nova_libvirt without stopping instances ๐Ÿ–ฅ๏ธ Single VM Restarting nova_libvirt does not stop instances
5 Restarting the Open vSwitch agent without cutting the network ๐Ÿ–ฅ๏ธ Single VM Restarting the OVS agent does not cut instances
6 An instance unreachable from outside ๐Ÿ–ฅ๏ธ Single VM Instance unreachable from outside
7 A compute goes down: fence it before you evacuate ๐Ÿ–ฅ๏ธ๐Ÿ–ฅ๏ธ๐Ÿ–ฅ๏ธ Multinode Fence it before you evacuate
8 Drain a compute for planned maintenance ๐Ÿ–ฅ๏ธ๐Ÿ–ฅ๏ธ๐Ÿ–ฅ๏ธ Multinode Moving instances off a host for maintenance
9 Live migration between the two computes ๐Ÿ–ฅ๏ธ๐Ÿ–ฅ๏ธ๐Ÿ–ฅ๏ธ Multinode Moving instances off a host for maintenance
10 Lose one controller: what keeps working ๐Ÿ–ฅ๏ธร—5 Multinode, 3 controllers Pattern (draft): three controllers, two computes
11 Lose two controllers: quorum lost, and recovered ๐Ÿ–ฅ๏ธร—5 Multinode, 3 controllers Procedure: MariaDB cluster recovery

Tip

In one minute (for managers)

  • What: a complete, disposable OpenStack for learning and for trying a procedure before production.
  • Cost: one run on the single VM measured about 0.05 EUR (0.19 TND) for 57 minutes. The script shows the cost per hour and asks first.
  • Risk: the VMs cost money every hour they exist. Use a prepaid cloud or a budget alert; one command deletes everything.
  • Who: anyone comfortable with a terminal and SSH; no OpenStack knowledge needed to start.
  • Not included: production use, support, block storage (Cinder), load balancing (Octavia); high availability only on the 5-VM multinode size.

๐Ÿ”ญ Going further

๐Ÿ‘ฅ Who this is for: newcomers who learn by doing, and operators who want a disposable lab to try a runbook. The newcomer path explains the vocabulary as you go.
๐Ÿ—๏ธ How the labs are built: the all-in-one lab pattern and the multinode pattern. Each guide states the releases its scripts were last run on.
๐Ÿ”’ This repository never contains credentials (clouds.yaml with secrets, passwords.yml) or marketing. Keep yours out of any fork or pull request.
๐Ÿค New variants, exercises and fixes are welcome: CONTRIBUTING.md. Any provider can be added as a variant on equal terms; nothing in the exercises depends on the provider.

๐ŸŒ Part of OzarLand Academy

This module is the first stop of OzarLand Academy, the learning path of OzarLand: a shared land for better OpenStack operations. After the lab, the path goes on with deployment patterns, day-2 operations with Kolla-Ansible, VM runbooks and the production guide. Questions, corrections and reports from your cloud are welcome in the Discussions.

โš–๏ธ License

Apache-2.0.