Bug 629217 - [RFE] Per user/group running recipes count quota
Summary: [RFE] Per user/group running recipes count quota
Keywords:
Status: CLOSED WONTFIX
Alias: None
Product: Beaker
Classification: Retired
Component: scheduler
Version: 0.5
Hardware: All
OS: Linux
low
medium
Target Milestone: ---
Assignee: beaker-dev-list
QA Contact:
URL:
Whiteboard: GroupModel
Depends On: 1127129
Blocks:
TreeView+ depends on / blocked
 
Reported: 2010-09-01 10:29 UTC by Marian Csontos
Modified: 2020-11-19 22:33 UTC (History)
5 users (show)

Fixed In Version:
Clone Of:
Environment:
Last Closed: 2020-11-19 22:29:36 UTC
Embargoed:


Attachments (Terms of Use)

Description Marian Csontos 2010-09-01 10:29:53 UTC
Description of problem:

Sometimes all the machines are blocked for a long time due to e.g.:
- bad habit of including reservesys at the end of all tasks
- people submitting same jobs multiple times
- people holding machines reserved
- or just too big load

It is often legal usage, so can not be banned/prevented in general.

We need to apply some level of self-control: I/we can not run more jobs: have to release some machines...

Solution(?):

Implement recipe quotas, so one user/group does not block all machines for a long time whatever their job priority is.

This would require a "sudo"/rungroup command to allow user to pick a group for a job if he is a member of multiple groups.

Comment 1 Nick Coghlan 2012-10-17 04:40:07 UTC
Bulk reassignment of issues as Bill has moved to another team.

Comment 4 Nick Coghlan 2014-08-06 08:07:58 UTC
The current scheduler isn't clever enough to have different quotas based on which systems they're using, but we could potentially limit access to the *general* pool this way.

Anything more sophisticated would need to wait until we're able to switch to a more sophisticated scheduling engine.

Comment 5 Nick Coghlan 2014-08-08 01:47:16 UTC
This is on hold until we evaluate the possibility of switching to a more capable scheduling engine.

Comment 6 Nick Coghlan 2014-08-08 01:50:13 UTC
If we decide to pursue a simpler scheme that only limits access to the general pool, that will be filed as a separate RFE based on usage analysis of Red Hat's main Beaker instance. It's a tricky problem, since preliminary checks of an OS build are *supposed* to spin up hundreds of recipes in a relative short span of time.


Note You need to log in before you can comment on or make changes to this bug.