Software Updates Overview
Overview of all AIsware software update procedures.
This document is the entry point for all AIsware software update procedures. It describes what components exist, how updates are discovered, and which updates require downtime.
Component Inventory
| Component | Update Frequency |
|---|---|
| AIsware PaaS (iaas-console / observability) | Each minor AIsware release |
| Management cluster | As needed; typically follows K3s releases |
| OpenStack | Each major AIsware release |
| Inventory / Configuration | As needed for config changes |
| Guest images | Each AIsware release |
| Node OS (Control and Compute Nodes) | Point releases, kernel security patches |
Discovery of Updates
Operators learn about new updates through:
- Release Notifications: Announcements sent to the
ai-factory-announcemailing list. - Release Notes: Published in the AIsware Documentation.
Distribution of Updates
Updates are distributed through the following channels:
- Release Bundles (Automation & Scripts):
- Distributed as downloadable archives (for example,
.tar.gz) linked in the Release Notes and Announcement emails. - Operators download and extract the bundle to the bastion host.
- Distributed as downloadable archives (for example,
- Container Images:
- Hosted on the project's Container Registry.
platform-setup.shpulls these images during the deployment phase.
- Guest Images:
- Hosted on the project's S3-like storage.
platform-setup.shdownloads these images during the deployment phase.
- Node OS Images (Control and Compute Nodes):
- Hosted on the project's S3-like storage.
- The PXE server downloads these images during the
platform-setup.sh --bootstrapexecution.
Downtime Matrix
| Update Type | Downtime? | Notes |
|---|---|---|
| AIsware PaaS (iaas-console / observability) | No downtime | Rolling K8s deployment; brief pod restart only |
| Management cluster update | Management plane only | OpenStack API and workloads unaffected; iaas-console & observability platform briefly unavailable |
| OpenStack version upgrade | Maintenance window required | Service restarts, haproxy brief outage; notify tenants |
| Configuration-only changes (inventory) | Depends on what changed | See CONFIGURATION for per-variable impact |
| Guest images update | No downtime | New images available immediately; existing instances unaffected |
| Node OS re-provisioning | Full node downtime | Loss of compute capacity on target node; evacuate Nova instances first |