Red Hat Bugzilla – Bug 814411
RFE: Add list of default groups in user preferences.
Last modified: 2015-09-13 22:03:10 EDT
Description of problem:
As of now, a user who is a member of multiple system groups has to explicitly exclude a group in their job xml to prevent machines in said group from picking up the job. This leads to a tedious and error prone process of having to manually edit the job xml and exclude the unwanted groups before submitting the job.
I would like to propose that we add a default groups list as one of the user preferences. If a job xml does not include any group specification then only the systems in the user's default groups will be used to run the job. By default we can include all the groups the user is a member of so that users who have no need for this feature can continue to work as they do today.
Alice has some jobs she wants to run on every RHEL 9.x tree as it lands. It makes no difference to her which systems run those jobs so there are no group specifications in any of her xml files.
Alice then goes on vacation for a month and her testing responsibilities transition to Bob during that time. Bob however is a member of the sooper-sekrit-hardware group which consists of 10 very rare prototype systems. Bob would like to save those systems for testing specific to that hardware and they are not for general use.
Bob currently has to modify all of Alice's xml files to exclude that group that he is in. It would be much easier for him to be able to exclude that group by default and only have jobs run on systems in that group if he specifies either a hostname in that group or the group itself in the job xml.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
This "Use only when explicitly requested" behaviour should be considered as a possible element in the System Pools feature planned for Beaker 1.1 (see http://beaker-project.org/dev/proposals/system-pools.html)
Hi Nick, any update on this potential feature?
Hi Evan, sorry for the delayed reply. There unfortunately aren't any near term changes planned in this area - marking systems as "Prototype" so they're automatically excluded by the default "Machine" filter remains the best option for ensuring they're only selected when you intended to use them. Another alternative these days, if only addressing them by name is an acceptable limitation, is to put them in Manual mode and only access them via forced identity-based scheduling rather than the regular host filtering mechanism.