Bug 1260381 - Incorrect behavior of power saving weight module
Summary: Incorrect behavior of power saving weight module
Keywords:
Status: CLOSED CURRENTRELEASE
Alias: None
Product: ovirt-engine
Classification: oVirt
Component: General
Version: 3.6.0
Hardware: All
OS: Linux
high
medium
Target Milestone: ovirt-4.0.2
: 4.0.2.1
Assignee: Yanir Quinn
QA Contact: Artyom
URL:
Whiteboard:
Depends On:
Blocks: 1349345
TreeView+ depends on / blocked
 
Reported: 2015-09-06 09:59 UTC by Artyom
Modified: 2016-08-12 14:29 UTC (History)
14 users (show)

Fixed In Version:
Clone Of:
: 1349345 (view as bug list)
Environment:
Last Closed: 2016-08-12 14:29:58 UTC
oVirt Team: SLA
Embargoed:
dfediuck: ovirt-4.0.z+
mgoldboi: planning_ack+
rgolan: devel_ack+
mavital: testing_ack+


Attachments (Terms of Use)


Links
System ID Private Priority Status Summary Last Updated
oVirt gerrit 57740 0 master MERGED core :fix PowerSavingWeight score calculation 2016-06-07 08:35:58 UTC
oVirt gerrit 59674 0 ovirt-engine-4.0 MERGED core: fix PowerSavingWeight score calculation 2016-07-12 10:03:17 UTC

Description Artyom 2015-09-06 09:59:40 UTC
Description of problem:
Power saving weight module not work in proper way for hosts with zero vms.

Version-Release number of selected component (if applicable):
rhevm-backend-3.6.0-0.12.master.el6.noarch

How reproducible:
Always

Steps to Reproduce:
1. Have two hosts with the same CPU arch(host_1, host_2)
2. Run vm with 1Gb on host_1
3. Change scheduler policy to power_saving with default parameters
4. Start additional vm

Actual results:
Vm start on host_2

Expected results:
Vm start on host_1

Additional info:
Problem that when host have zero vms, engine give to host score MaxSchedulerWeight - 1(999), but hosts with vms can have score bigger than(999) because expression:
"score -= vds.getMaxSchedulingMemory() / 100;" under calcEvenDistributionScore method in EvenDistributionWeightPolicyUnit class, we can have have negative score, and then under score method in PowerSavingWeightPolicyUnit class we run "score -= calcEvenDistributionScore(vds, vm, vdsGroup != null ? vdsGroup.getCountThreadsAsCores() : false);"
we can have have score bigger than MaxSchedulerWeight score.

Comment 2 Roy Golan 2015-12-16 09:55:29 UTC
Possibly a bug. Martin will take a look.

Comment 4 Red Hat Bugzilla Rules Engine 2016-07-03 10:37:19 UTC
Target release should be placed once a package build is known to fix a issue. Since this bug is not modified, the target version has been reset. Please use target milestone to plan a fix for a oVirt release.

Comment 5 Sandro Bonazzola 2016-07-12 10:05:12 UTC
Moving back to POST since this bug has been targeted to 4.0.1 but referenced patches are not included in 4.0.1 branch.
Please check this bug status and eventually re-target to 4.0.2.

Comment 6 Artyom 2016-07-24 15:04:06 UTC
Verified on rhevm-4.0.2-0.1.rc.el7ev.noarch


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