Bug 1127129

Summary: [RFE] Switch to a more sophisticated scheduling engine
Product: [Retired] Beaker Reporter: Nick Coghlan <ncoghlan>
Component: schedulerAssignee: beaker-dev-list
Status: CLOSED WONTFIX QA Contact: tools-bugs <tools-bugs>
Severity: unspecified Docs Contact:
Priority: unspecified    
Version: developCC: cbouchar
Target Milestone: ---Keywords: FutureFeature
Target Release: ---   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard:
Fixed In Version: Doc Type: Enhancement
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2020-11-12 20:31:27 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Bug Depends On:    
Bug Blocks: 574930, 598781, 629155, 629217, 709012, 983037, 1212849    

Description Nick Coghlan 2014-08-06 08:34:32 UTC
Currently, Beaker uses a fairly simply custom scheduling system that repeatedly scans over all queued recipes and tries to assign each one to a currently free host.

Some RFEs are currently impossible to implement given the simplicity of the scheduling model.

This RFE is being added in order to more easily capture which bugs are blocked by that.

One proposal to improve the situation is captured in https://beaker-project.org/dev/proposals/event-driven-scheduler.html

As noted in the deferral of that proposal, another option is to switch away from running a custom scheduler, and instead adopt the Apache Mesos meta-scheduling framework for Beaker's resource management: https://mesos.apache.org/

The latter approach seems desirable (doing scheduling well is difficult, and Mesos provides a lot of useful infrastructure), but what Beaker does is a little odd from the perspective of most resource management systems (we don't just need "a computer", we need to get very, very specific, down to the level of identifying specific models of harddrive controller, or network card).

This is currently stalled on lack of time to continue the investigation.

Comment 1 Nick Coghlan 2014-08-06 08:38:55 UTC
Ray, could I get you to provide a summary of where you got to with your research here?

https://lists.fedorahosted.org/pipermail/beaker-devel/2014-May/000997.html was my decision to defer the existing design proposals, but I didn't publish a lot in the way of technical details.

Comment 3 Roman Joost 2016-06-07 05:44:13 UTC
Since Raymond departed, I think I can at least clear the NEEDINFO flag.