Skip to content
Open
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 18 additions & 2 deletions docs/deepops/update-deepops.md
Original file line number Diff line number Diff line change
Expand Up @@ -538,10 +538,26 @@ ansible-playbook -l slurm-node playbooks/slurm-cluster/nvidia-dcgm-exporter.yml
To update Pyxis and/or Enroot, edit your DeepOps configuration and specify the new versions you wise to use:

```bash
slurm_pyxis_version: "0.11.1"
enroot_version: "3.2.0"
slurm_pyxis_version: "0.24.0"
enroot_version: "4.2.1"
```

On RHEL-family hosts, override the RPM URLs alongside the version. The
`nvidia.enroot` role builds them with a hardcoded `el7` tag, but Enroot moved
to `el8` artifacts in 3.4.1 and stopped publishing `el7` ones, so raising
`enroot_version` to 3.4.1 or later on its own makes the download 404:

```bash
enroot_rpm_packages:
- 'https://github.com/NVIDIA/enroot/releases/download/v{{ enroot_version }}/enroot-{{ enroot_version_string }}.el8.x86_64.rpm'
- 'https://github.com/NVIDIA/enroot/releases/download/v{{ enroot_version }}/enroot+caps-{{ enroot_version_string }}.el8.x86_64.rpm'
```

One version needs more than that: 3.4.0 ships `enroot_3.4.0-1_amd64.deb`
alongside `enroot-3.4.0-2.el7.x86_64.rpm`, and the role derives both filenames
from a single `enroot_release`, so no value of it satisfies both package
families. It is the only release where the two disagree; pick a different one.

Then re-run the Pyxis playbook:

```bash
Expand Down