Bug 722321

Summary:	Beaker scheduler not honoring shared vs. unshared, kicking off installs on machines that it shouldn't be
Product:	[Retired] Beaker	Reporter:	Derrich Hafemann <dhafeman>
Component:	scheduler	Assignee:	Bill Peck <bpeck>
Status:	CLOSED CURRENTRELEASE	QA Contact:
Severity:	urgent	Docs Contact:
Priority:	unspecified
Version:	0.6	CC:	bpeck, dcallagh, mcsontos, rmancy, stl
Target Milestone:	---
Target Release:	---
Hardware:	All
OS:	Linux
Whiteboard:
Fixed In Version:		Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2011-07-28 23:50:11 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:

Description Derrich Hafemann 2011-07-14 21:34:26 UTC

Description of problem:

When using the reserve workflow to find any old machine I could install Fedora on, Beaker scheduled jobs to run against systems that were set as unshared. This should never happen.

Version-Release number of selected component (if applicable):

Scheduler component in release 0.6.15-1

How reproducible:

Honestly, I'm a bit afraid to test and find out. I'd expect 100%, but I don't want to clobber any more systems than I may have.


Steps to Reproduce:
1. Use the Reserve Workflow, specify Fedora 14 (or Fedora 15), STABLE and the NFS variant
2. Enter any text for whiteboard, submit job.
3. View the Beaker page for the system that was chosen for the job

Actual results:
In my cases (jobs 111806, 111807), both auto-picked systems were not set to shared.


Expected results:

Beaker should only schedule jobs against systems which have been configured to be used as test clients.


Additional info:

Once I saw which systems had been picked for the jobs, I cancelled them as quickly as I could. I don't know how far along the installs got, but I hope it wasn't far enough to write anything to the disk.

Comment 1 Bill Peck 2011-07-14 22:01:38 UTC

from irc

<bpeck> its probably because your an admin
 we have access to *ALL* systems
 I usually add this to my recipes
 <group op="=" value=""/>
 in HostRequires
 that keeps the query to only systems that are shared to everyone

Comment 2 Derrich Hafemann 2011-07-14 22:09:03 UTC

Regardless of whether the user submitting the job is an administrator or not, shouldn't the scheduler honor whether a given system has been configured as an automated test system or not?

Comment 3 Dan Callaghan 2011-07-14 22:36:20 UTC

(In reply to comment #2)

If the system should have no automated jobs run on it by the scheduler, its condition should be set to "Manual" (rather than "Automated").

Comment 4 Bill Peck 2011-07-18 14:17:24 UTC

After discussing this with Derrich I believe it makes sense to remove the super user scheduling ability of the admin group.

If an admin needs to schedule against a system that he doesn't have access to he can always loan it to himself.

Comment 5 Bill Peck 2011-07-21 18:06:40 UTC

code has been pushed to gerrit.

unit tests created for beakerd and webUI.