[Beowulf] Your thoughts on the latest RHEL drama?

[Roland Fehrenbacher]

Hi John, Tony,

On 6/28/23 10:18, Tony Travis wrote:
> On 28/06/2023 07:18, John Hearns wrote:
>> Rugged individuaiist? I like that...    Me puts on plaid shirt and 
>> goes to wrestle with some bears,,,
>>
>>  > Maybe it is time for an HPC Linux distro, this is where
>> Good move. I would say a lightweight distro that does not do much nd 
>> is rebooted every time a job finishes.
>> Wonder what security types would think of that....
>>
>> Sidelining the discussion a bit I have been involved with projects 
>> where security types insist on the entire stack for firmware upwards 
>> is kept up to date.
>> This feeds into the Redha debate of course - if we go Debian how do 
>> you satisfy corporate types?
>> i guess ubuntu has a role here.
>
> Hi, John.
>
> There is already an Ubuntu-based HPC distro: Qlustar
>
>> https://qlustar.com/

let me chime in here and explain a little about Qlustar. While it is 
certainly a distro (actually two, one based on Ubuntu, the other one on 
RHELcompat  - Alma 8 and CentOS 7), it is much better viewed as a 
ClusterOS for HPC/AI and Storage. The architecture is such that cluster 
head-nodes always run on Ubuntu LTS (Qlustar 13 <-> Ubuntu 22.04, 
Qlustar 12 <-> Ubuntu 20.04), but nodes currently may also run Alma 8 
and CentOS 7.

Anything other than a head-node boots images via the net and images are 
sent in a two-stage process (small initrd via normal PXE < 40MB, which 
pulls the real squashfs image via a custom multicast client).

HPC stuff like ready-to-run up-to-date Nvidia drivers, MLNX OFED, Slurm, 
Lustre, Spack, BeeGFS, ... is all integrated and working out-of-the-box. 
Security updates for all this stuff typically coming like every 6 weeks. 
No more worries about version changes between all these components, 
Qlustar takes care of this.

To manage the cluster beast there is a constantly improving management 
framework: QluMan (for detailed feature description have a look at its 
manual at 
https://docs.qlustar.com/Qlustar/13/ClusterOS/qluman-guide/Introduction.html)

Actually something like an upgrade path for people wanting to migrate 
away from RHELcompat to Ubuntu is rather smooth:

 1. Install a Qlustar head.
 2. Setup some nodes with CentOS 7 or Alma 8 and the required RPMs to
    match your earlier environment.
 3. Have other nodes running Ubuntu.
 4. Setup some slurm queues for the different archs.
 5. Your users should then be able to run their codes on the RHELcompat
    nodes and can get started to port things to Ubuntu at the same time.
    If you're using spack e.g., this should really be a piece of cake.

Our software is 100% open source, will stay that way forever and yes, 
there are many clusters worldwide running on Qlustar, academia, research 
labs, commercial. Qlustar is financed via support contracts sold by 
Q-Leap, no external invest money, so no external interests to mess 
things up.

I kind of hate to send messages like this that might come across like 
marketing. On the other hand, I feel that the community doesn't quite 
know well enough, what Qlustar can provide and needs some more 
info/assurance/guidance before making the big step with migration away 
from the known and trusted. In any case, I felt since quite a while that 
more community building from our side is necessary and maybe this is a 
good moment to start.

Ready for more discussion,

Roland

-------
https://qlustar.com
   -- 100% Open Source HPC / Storage / Cloud Linux Cluster OS --
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://beowulf.org/pipermail/beowulf/attachments/20230628/af13fc8d/attachment.htm>