[slurm-users] Federated SLURM

[Kees de Jong]

Hi,



I've been tinkering around with federated SLURM and I have a few
questions regarding how federated it really can be.


* A federated cluster can consist out of clusters using SLURM > 17.11,
these clusters then share the *same* database, correct?
* Mixing different SLURM versions with a slurmdbd is not supported,
correct?
* Using this illustration [1], a federated cluster must use the same
SLURM version as the "on-premise" cluster slurmdbd version, which runs
the central database. Thus the Google Cloud Platform must run the same
SLURM version as the "on-premise" one for compatibility, correct?


Other concerns are e.g. different mounted file systems (Lustre), and
other differences.

If SLURM versions must be compatible and jobs also expect a compatible
system (if cloud bursting is applied with large jobs). So the main
question then becomes; how federated can federated SLURM be? It seems
to be that identical sites are required in order to make this work.

With the definition of federation, I was expecting to a certain extend
autonomous sites, joined together in a federation. If I'm right or
wrong, please let me know! :)


[1] 
https://codelabs.developers.google.com/codelabs/hpc-slurm-federated-on-gcp/#0



-- 
Met vriendelijke groet,
Kees de Jong

De informatie opgenomen in deze e-mail kan vertrouwelijk zijn en is
uitsluitend bestemd voor de geadresseerde(n). Indien u deze e-mail
onterecht ontvangt, wordt u verzocht de inhoud niet te gebruiken en de
afzender direct te informeren door de e-mail te retourneren. Aan deze
e-mail inclusief de bijlagen kunnen geen rechten ontleend worden,
tenzij schriftelijk anders wordt overeengekomen.
--
The information contained in this e-mail may be confidential and is
intended to be exclusively for the addressee(s). Should you receive
this e-mail unintentionally, please do not use the contents herein and
notify the sender immediately by return e-mail. This e-mail including
the attachments are not legally binding, unless otherwise agreed upon
in writing.
--
OpenPGP fingerprint: 0x0E45C98AB51428E6
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 833 bytes
Desc: This is a digitally signed message part
URL: <http://lists.schedmd.com/pipermail/slurm-users/attachments/20191214/839cad4f/attachment-0001.sig>