Skip to content

B OpenNebula/engineering#976: Fix on-premises AI Factory deployment guide - #1082

Merged
rsmontero merged 1 commit into
one-7.4from
b-engineering-976
Oct 29, 2026
Merged

rsmontero merged 1 commit into
one-7.4from
b-engineering-976

Conversation

@mattrowe-opennebula

@mattrowe-opennebula mattrowe-opennebula commented Oct 28, 2026 •

Copy link
Copy Markdown
Contributor

Description

Fix the on-premises AI Factory deployment guide so its GPU inventories bind the devices to vfio-pci and include them in the Host's PCI filter with current OneDeploy. The guide also needs Front-end hostname resolution and passwordless root SSH prerequisites, and should use Hatch's Ansible environment consistently.

  • Add set_driver: vfio-pci and unlisted: false to all four GPU entries, document their defaults, and show full-domain PCI addresses with lspci -Dnn.
  • Explain hostname resolution, ensure_hosts and hostname renaming; move root SSH setup before both deployment examples.
  • Remove the separate pipx Ansible installation and make hatch shell optional.
  • Correct AMD IOMMU guidance in the deployment page and linked Host Configuration guide; make bootloader instructions visible for every architecture.
  • Replace the stale OneDeploy PCI wiki link, repair the single-Host section link, and document the non-GPU node's default filtering behavior.

Fixes OpenNebula/engineering#976. Based on the supplied deployment report dated 2026-10-28.

Validation: source review against OneDeploy 386aed1 and Linux 6.8 kernel documentation; checked all four GPU entries, modified links/anchors, and git diff --check. No builds or deployment commands were run.

Documentation-only scope: no release-note changes. After merge, cherry-pick to master and one-7.4-maintenance.

Branches to which this PR applies

  • master
  • one-7.4
  • one-7.4-maintenance

  • Check this if this PR should not be squashed

@rpetrescu-one rpetrescu-one left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

lgtm

@rsmontero
rsmontero merged commit 8766b69 into one-7.4 Oct 29, 2026
1 check passed
@rsmontero
rsmontero deleted the b-engineering-976 branch October 29, 2026 09:10
rsmontero pushed a commit that referenced this pull request Oct 29, 2026
…uide (#1082)

Fix the on-premises AI Factory deployment guide so its GPU inventories
bind the devices to `vfio-pci` and include them in the Host's PCI filter
with current OneDeploy. The guide also needs Front-end hostname
resolution and passwordless root SSH prerequisites, and should use
Hatch's Ansible environment consistently.

- Add `set_driver: vfio-pci` and `unlisted: false` to all four GPU
entries, document their defaults, and show full-domain PCI addresses
with `lspci -Dnn`.
- Explain hostname resolution, `ensure_hosts` and hostname renaming;
move root SSH setup before both deployment examples.
- Remove the separate pipx Ansible installation and make `hatch shell`
optional.
- Correct AMD IOMMU guidance in the deployment page and linked Host
Configuration guide; make bootloader instructions visible for every
architecture.
- Replace the stale OneDeploy PCI wiki link, repair the single-Host
section link, and document the non-GPU node's default filtering
behavior.

(cherry picked from commit 8766b69)
rsmontero pushed a commit that referenced this pull request Oct 29, 2026
…uide (#1082)

Fix the on-premises AI Factory deployment guide so its GPU inventories
bind the devices to `vfio-pci` and include them in the Host's PCI filter
with current OneDeploy. The guide also needs Front-end hostname
resolution and passwordless root SSH prerequisites, and should use
Hatch's Ansible environment consistently.

- Add `set_driver: vfio-pci` and `unlisted: false` to all four GPU
entries, document their defaults, and show full-domain PCI addresses
with `lspci -Dnn`.
- Explain hostname resolution, `ensure_hosts` and hostname renaming;
move root SSH setup before both deployment examples.
- Remove the separate pipx Ansible installation and make `hatch shell`
optional.
- Correct AMD IOMMU guidance in the deployment page and linked Host
Configuration guide; make bootloader instructions visible for every
architecture.
- Replace the stale OneDeploy PCI wiki link, repair the single-Host
section link, and document the non-GPU node's default filtering
behavior.

(cherry picked from commit 8766b69)
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

3 participants