Small configuration issues (slurm, chrony, docker image)

[Rknegt]

Hi,

During the further installation of TrinityX I came across some issues that I wanted to let you know. If you would prefer these issues to be listed as separate issues, let me know.

1. Multiple controllers using SlurmctldHost

Currently, in a 2 controller setup Slurm is configured like:

trinityX/site/roles/trinity/slurm/templates/slurm.conf.j2

Line 4 in dd23d18

SlurmctldHost={{ slurm_ctrl_list }}

SlurmctldHost=controller1,controller2

However, multiple controllers should be configured on its own line:
https://slurm.schedmd.com/slurm.conf.html#OPT_SlurmctldHost

SlurmctldHost=controller1
SlurmctldHost=controller2

I found this issue during debugging an DNS issue and saw that there were queries done for the hostname: controller1,controller2.

2. chrony.conf.j2 word swap

When configuring allowed networks for chrony I came across another bug. I think the one of {{ net }}, {{ server }} variables should be swapped in this case.

trinityX/site/roles/trinity/chrony/templates/chrony.conf.j2

Line 25 in dd23d18

{% for server in chrony_allow_networks %}

{% if chrony_allow_networks %}
{% for server in chrony_allow_networks %}
allow {{ net }}
{% endfor %}

3. Docker images with slash (rockylinux/rockylinux:9.7)

When trying to install any recent Rocky Linux 9.3+ releases via docker we need to use the rockylinux/rockylinux images. The issue that there is a slash in the container name breaking the Ansible code because the folder doesn't exist and need to be created.

After creating /trinity/downloads/docker-image-rockylinux, the Rocky 9.7 does work apart from one small issue of installing the selinux-policy-minimum rpm that is not available anymore on RHEL/Rocky 9+.

trinityX/site/roles/trinity/image-download/tasks/main.yml

Line 59 in dd23d18

shell: "docker export {{ docker_container_id.stdout }} > {{ trix_root }}/downloads/docker-image-{{ image_download_distribution }}.tar"

4. Cannot change tuned profile using tuned-adm in lchroot

I found that setting the compute performance tuned profile fails in building the compute default images. At least for Rocky 9.3:

TASK [trinity/tunables : Setting compute performance tuned profile] **************************************************************************************************************************
2026-04-22 16:24:57,022 p=1192670 u=root n=ansible | fatal: [compute.osimages.luna]: FAILED! => {"changed": true, "cmd": "tuned-adm profile hpc-compute", "delta": "0:00:00.159230", "end": "2026-04-22 16:24:56.979824", "msg": "non-zero return code", "rc": 1, "start": "2026-04-22 16:24:56.820594", "stderr": "Cannot talk to TuneD daemon via DBus. Is TuneD daemon running?\nTuneD (re)start failed, check TuneD logs for details.\nUnable to switch profile.", "stderr_lines": ["Cannot talk to TuneD daemon via DBus. Is TuneD daemon running?", "TuneD (re)start failed, check TuneD logs for details.", "Unable to switch profile."], "stdout": "Trying to (re)start tuned...", "stdout_lines": ["Trying to (re)start tuned..."]}

Probably this could be solved with setting it using the file.

echo "hpc-compute" > /etc/tuned/active_profile

Thanks,
Rick

[jw-26-sudo]

Hello everybody,

With reference to recent/previous issues regarding TrinityX setup with SLURM, should we setup SLURM first then we can proceed to setup TrinityX?

If anyone has experience or a working method for setting up TrinityX with SLURM, it will be best to provide the steps/references/workarounds.

Thanks in advance!

Regards,
Jing Wei

[aphmschonewille]

First i want to thank Rick for his input.
Where did not change this yet for upcoming release, we note. Not all things seem to have krept up and causing issues with recent ansible, some can be changed, some already have but not released yet.

To Jing, slurm will be automatically confirgured by Trinityx out-of-box the box.

[jw-26-sudo]

Hi Antoine,

Appreciate if the following sentence can be elaborated further:

Slurm will be automatically configured by TrinityX out-of-box the box.

Thanks you.

Regards,
Jing Wei

[aphmschonewille]

Slurm is integral with TrinityX installation. Configuration is also done.
See here as well: https://docs.clustervision.com/admin/wlm/

[Rknegt]

First i want to thank Rick for his input. Where did not change this yet for upcoming release, we note. Not all things seem to have krept up and causing issues with recent ansible, some can be changed, some already have but not released yet.

To Jing, slurm will be automatically confirgured by Trinityx out-of-box the box.

Thanks for the update. Feel free to close the issue if the things are already picked up or noted down somewhere else.

Rick

[jw-26-sudo]

Slurm is integral with TrinityX installation. Configuration is also done. See here as well: https://docs.clustervision.com/admin/wlm/

Hi Antoine,

Are you able to advise on the below screenshot from my head node:

After running ansible-playbook controller.yml --tags slurm (if want to set up SLURM after TrinityX installation) from my head node, do we need to check and edit the following files:

1. /etc/slurm/slurm-nodes.conf
2. /etc/slurm/slurm-partitions.conf

Should I consider creating an inventory file in ~/trinityX/site folder as well?

And when I run "scontrol show nodes" from my head node, it shows "No nodes in the system". Do we need to provision the compute nodes first?

Also understand that the following regarding current Ubuntu OS compute image creation and deployment:

1. Slurm is not supported.
2. Diskless mode is working but not diskfull mode.

Sorry, because I saw this thread is now open. If you would prefer me to open up a new thread, please let me know.

Thank you.

Regards,
Jing Wei

[aphmschonewille]

Hi Jing Wei,

Indeed for 15.3 ubuntu comes without slurm support due to unavailability of deb packages in the openhpc stack. This however will be replaced by our own slurm packages to support all major (incl ubuntu) distributions, coming in release 16. We are also looking into offering support for diskfull support for ubuntu as well.

It wasn't clear to me that you were using ubuntu and i assumed a redhat based image. In the past most cases for ubuntu were k3s, however we have noticed a shift towards hpc utilization as well. I hope that you can wait for 16, due mid of june, or that will do an upgrade once available.

-Antoine