Kolla-Ansible Production Architecture
How Kolla-Ansible actually structures an OpenStack deployment, where configuration comes from, and how to make changes without discovering them during an outage.
expert· 3 minOpenStackKolla-Ansible
Cloud topic
OpenStack
OpenStack is well documented as a set of projects and poorly documented as a system you have to keep running. This hub is about the second thing: how the services depend on each other, how they fail quietly, and what to check first when something that should be simple is not working.
The services, and how each one fails
Most OpenStack incidents are not the service you are looking at. They are a dependency of that service that stopped participating without going down.
| Service | Responsibility | Characteristic failure |
|---|---|---|
| Keystone | Identity, tokens, service catalog | Token validation failures cascade into every other service at once. |
| Nova | Compute scheduling and instance lifecycle | NoValidHost when the scheduler has no candidates — often a service-state problem. |
| Neutron | Networking, ports, routers, security groups | An agent goes down and instances build but never gain connectivity. |
| Cinder | Block storage and volume scheduling | Volume service stops reporting capacity; every create fails scheduling. |
| Glance | Image storage and metadata | Slow or failing image downloads present as slow instance builds. |
| Placement | Resource inventory and allocation | Stale allocations make capacity look consumed when it is free. |
| RabbitMQ | Message bus between all services | Partitions leave services running but no longer talking to each other. |
| MariaDB / Galera | Persistent state for every service | Cluster loses quorum; writes fail while reads appear healthy. |
First five minutes
Read state and status as two separate columns
Every OpenStack service listing reports administrative intent and actual state separately. enabled means someone wants it to run. up means it reported in recently. A service that is enabled and down is not a configuration question — it is the incident.
$ openstack compute service list$ openstack network agent list$ openstack volume service list# Then, only if those are clean:$ openstack hypervisor list$ openstack resource provider list# Placement disagreeing with Nova about capacity is a# classic source of phantom "no capacity" failures.
Guides
OpenStack guides
OpenStack Production Operations
Operating OpenStack in production: service state versus status, the message bus, placement disagreements, and the quiet failures that keep dashboards green.
expert· 3 minOpenStackNova
Runbooks
OpenStack runbooks
Procedures with preconditions, validation, and rollback — written to be followed under pressure.
Cinder — No Valid Backend
Diagnose and remediate OpenStack volume creation failures caused by the scheduler having no candidate backends, including the case where the API reports healthy.
expert· 20 minOpenStackCinder
Newsletter
Inside The AI Stack Brief
A practical weekly briefing on AI engineering, infrastructure, production operations, and the technologies powering the AI stack.
One email a week. No sponsorship placements inside the technical sections. Unsubscribe in one click.