Ubuntu diskfull support for TrinityX 16

[jw-26-sudo]

Hi Antoine,

I was referring to the TrinityX 16 release regarding Ubuntu diskfull support.

I have reviewed the currently available resources and solutions. From my understanding, the existing implementations appear to support only Red Hat and SUSE family images.

Could you please advise whether we should wait for the enhanced diskfull and RAID deployment scripts to be extended to Ubuntu, or if there are any recommended approaches available at the moment?

Thank you for your guidance.

Regards,
Jing Wei

[aphmschonewille]

Hi Jing Wei,

Trix 16 has full support using the provided diskfull and raid1 methods. These are out-of-the-box methods, but of course the scripts are free to be modified to your needs, e.g. raid0 or raid5, etc.

Note that we are currently evaluating options to make this even easier, but that wasn't ready for the latest release. I wouldn't worry too much about it, as there will always be an upgrade path.

For now, if you want to use ubuntu with the provided scripts, I'd say go for it.

-Antoine

[jw-26-sudo]

Hi Antoine,

After the TrinityX 16 release it will be very easy to include just support for ubuntu disk scripts: ansible-playbook controller.yml --tags=luna. After this step, ensure script is set for node/group following the online documentation and it should work out of the box.

So if I run ansible-playbook controller.yml, it will support diskfull for Ubuntu, RHEL, and SUSE family OS images?

Note that based on the documentation, we can use the following valid script plugin names: diskfull and raid1.

In addition, I just want to confirm if there are outstanding SLURM issues to be resolved, such as slurmd and slurmdbd? Are these expected to be released in TrinityX 16.1?

Thank you.

Regards,
Jing Wei

[jw-26-sudo]

Hi Antoine,

I hope you're having a good day.

I just wanted to follow up on my previous post regarding diskfull and SLURM issues. I would appreciate any updates when convenient.

Thank you, and I look forward to your response.

Best regards,
Jing Wei

[jw-26-sudo]

Hi Antoine,

I am trying to set up a TrinityX version 16 (HA cluster) using Rocky version 9.8 via Proxmox VMs.

I ran ansible-playbook controller.yml on primary node, then secondary node. So will it support diskfull only for RHEL and SUSE family OS images? I understood that if I run ansible-playbook controller.yml --tags=luna, it will support diskfull only for Ubuntu family OS images. Please correct me if I misunderstood anything.

I am currently trying to create an Ubuntu 24 OS compute image with the following:

1. alternative_distribution: Ubuntu-24
2. alternative_image_source: 'base'

But I have encountered the error on the below screenshot.

Can I conclude if this has got to do with me running only ansible-playbook controller.yml, instead of ansible-playbook controller.yml --tags=luna? So in short, ansible-playbook controller.yml does not support diskfull for Ubuntu family OS images?

In addition, after successfully creating an Ubuntu OS compute image, are we expecting the grub.cfg file by doing the following:

1. lchroot <name_of_created_ubuntu_image>
2. cd /boot/grub
3. We may be expecting the grub.cfg file inside the directory.

What if we are expecting the grub.cfg file inside the directory but the file does not generate/exist?

Finally, please advise if we should use the Ubuntu OS group name for the nodes in the TrinityX portal.

I appreciate if there are some better ways going forward.

Thank you.

Regards,
Jing Wei

[jw-26-sudo]

Hi Antoine,

I have managed to resolve the issue by using HTTP instead of HTTPS on the ~/trinityX/site/roles/trinity/image-create/vars/x64/image-ubuntu.yaml file.

Note that when I lchroot to my successfully created Ubuntu OS compute image, and went inside to /boot/grub folder, there is no grub.cfg file generated.

May I know if this is expected?

Finally, please advise:

1. if ansible-playbook controller.yml will support diskfull for RHEL and SUSE family OS images, and also for Ubuntu.
2. if we should use the Ubuntu OS group name for the nodes in the TrinityX portal.

Thank you.

Regards,
Jing Wei

[jw-26-sudo]

Hi Antoine,

During provisioning of the compute node using Ubuntu 24 OS diskfull install script, I encountered the following issue on the below screenshot:

Please advice on how to proceed further.

Thank you.

Regards,
Jing Wei

[jw-26-sudo]

Hi Antoine,

I have tried disabling DHCP on my compute node's internal network interface, and switching between "compute-group" and the created ubuntu OS group, but I encountered the following issue on the below screenshot (cannot resolve my floating hostname and IP >> trix_ctrl_ip and trix_ctrl_hostname):

The above error occurs when I tried the following:

1. Diskless and compute-image group
2. Diskless and ubuntu-image group
3. Diskfull and compute-image group
4. Diskfull and ubuntu-image group

Please advice on how to proceed further like:

1. running ansible-playbook controller.yml --tags=luna instead.
2. manually generating grub.cfg file and installing isc-dhcp-client inside the ubuntu OS image.
3. correct compute group name to use.

Thank you.

Regards,
Jing Wei

[jw-26-sudo]

Hi Antoine,

I am able to do a "diskfull" node provisioning using a Rocky OS compute image.

So for ansible-playbook controller.yml, does it require a "careful" project requirements gathering/planning?

The above question is based on my conclusion derived, where you can correct me if necessary:

• ansible-playbook controller.yml >> supports diskfull for RHEL and SUSE family OS images
• ansible-playbook controller.yml --tags=luna >> supports diskfull for Ubuntu family OS images only.

Thank you.

Regards,
Jing Wei

[jw-26-sudo]

Hi Antoine,

These are the things I have tried out today after running ansible-playbook controller.yml on my existing HA cluster:

1. ran ansible-playbook controller.yml --tags=luna
2. recreate Ubuntu version 24 OS compute image >> lchroot to ubuntu os image >> no grub.cfg file generated inside /boot/grub/ directory.
3. I was not able to achieve diskless and diskfull, facing the same errors using the Ubuntu OS compute image.
4. I was able to use a cloned Rocky OS compute image and achieve diskfull.

Appreciate if I can be enlightened on how to achieve diskfull for Ubuntu, RHEL and SUSE OS family images.

Thank you.

Regards,
Jing Wei

[jw-26-sudo]

Hi Antoine,

I hope you're having a good day.

Following up on my previous post(s) regarding Ubuntu diskfull/diskless issues, the same issues occur when I created an Ubuntu OS compute image with the following:

alternative_distribution: Ubuntu-22
alternative_image_source: 'base'

Issues I have reported previously:

1. No grub.cfg file generated inside /boot/grub directory after lchroot to <created_ubuntu_osimage>
2. /usr/sbin/dhclient: No such file or directory
3. Cannot resolve my floating hostname and IP >> trix_ctrl_ip and trix_ctrl_hostname) for HA cluster. This occurs likewise for non-HA cluster using trix_ctrl1_ip and trix_ctrl1_hostname. The IP I have used for both (non-HA and HA) clusters is an internal IP address.

Appreciate if you can provide some workaround solutions.

Thank you.

Regards,
Jing Wei

[jw-26-sudo]

Hi Antoine,

Are you able to advise if the following error is due to internal IP address out of range?

Cannot resolve my floating hostname and IP >> trix_ctrl_ip and trix_ctrl_hostname) for HA cluster. This occurs likewise for non-HA cluster using trix_ctrl1_ip and trix_ctrl1_hostname. The IP I have used for both (non-HA and HA) clusters is an internal IP address.

Thank you.

Regards,
Jing Wei

[jw-26-sudo]

accidentally closed this issue, now reopened it

[jw-26-sudo]

Hi Antoine,

I may have seen previous post(s) regarding setting up TrinityX cluster using Ubuntu OS.

I hope you can enlightened me with some points to take note of.

Just out of curiosity though, will things get complicated if I set up TrinityX cluster on Rocky/RHEL head node(s), then provision compute node(s) using an Ubuntu OS compute image?

Thank you.

Regards,
Jing Wei