[HTCondor-users] How to enforce user diversity per node?

[Thomas Hartmann <thomas.hartmann@xxxxxxx>]

Hi all,

we have been observing several pool nodes crashing [1] and we assume
that the crash pattern correlates with the overwhelming population of a
node with particular users.

So, we are thinking about how to foster the diversity of users per node.
I.e., how to let a node preferably request jobs from other users/groups,
if its currently allocated job slots belong 'mostly' to one user/group -
except if there are no other user/groups waiting or the nominal share
would be highly unfavourable?

Maybe somebody has already a recipe for something like that?

Cheers,
  Thomas


[1]
the FUSE kernel module crashed - apparently correlated with the number
of open file handles via fuse

Attachment: smime.p7s
Description: S/MIME Cryptographic Signature