[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Condor-users] Configuration
- Date: Tue, 02 Nov 2004 13:39:18 -0600
- From: Dan Bradley <dan@xxxxxxxxxxxx>
- Subject: Re: [Condor-users] Configuration
Support for this sort of group sharing policy is in active development.
In lieu of the full solution, there are various ways to get almost what
you want, with some undesired consequences.
One way is to pre-assign 100 machines to the group so that their jobs
only run there. Example:
#In the global condor_config
START = (...) && (TARGET.Group =?= UNDEFINED || MY.Group =?= TARGET.Group)
STARTD_EXPRS = Group
SUBMIT_EXPRS = Group
#In the local config file on the 100 machines:
Group = "OSG"
Depending on whether you also want them to have absolute (preemption)
priority on those machines, you can also set the machine RANK expression:
RANK = TARGET.Group =?= MY.Group
The main disadvantage to this setup is that the 100 nodes are hard-wired
rather than being chosen opportunistically.
Joseph L. Kaiser wrote:
I want to do the following:
I have a group of users who should, combined, only receive about 100
virtual CPU's out of a possible 272. If only one of these users in the
group is running jobs, they should be able to get all 100 CPU's. If
someone from that same groups starts running jobs, the first user's
number should decrease and the second user's number should increase but
they both shouldn't ever go about the 100 CPU's. Does anyone have an
idea on how I go about doing this?