Back to bug 1030064

Who When What Removed Added
Eoghan Glynn 2013-11-13 20:50:10 UTC Status NEW ASSIGNED
Assignee rhos-maint eglynn
Eoghan Glynn 2013-11-13 20:50:51 UTC Summary no autoscaling action occurs for percentage adjustment, depending on initial size & adjustment step size [heat] no autoscaling action occurs for percentage adjustment, depending on initial size & adjustment step size
RHEL Program Management 2013-11-13 20:57:34 UTC Target Release --- 4.0
Eoghan Glynn 2013-11-13 21:33:38 UTC Link ID OpenStack gerrit 56281
Steven Dake 2013-11-14 04:32:49 UTC Priority unspecified medium
Severity unspecified high
Eoghan Glynn 2013-11-14 07:34:35 UTC Status ASSIGNED POST
Eoghan Glynn 2013-11-14 10:53:53 UTC Target Milestone --- rc
Link ID OpenStack gerrit 56360
Steven Dake 2013-11-27 14:59:13 UTC Keywords OtherQA, Triaged
Priority medium high
QA Contact ajeain kwhitney
Eoghan Glynn 2013-12-04 15:31:28 UTC Link ID OpenStack gerrit 16394
Eoghan Glynn 2013-12-05 18:51:41 UTC Status POST MODIFIED
Fixed In Version openstack-heat-engine-2013.2-2.0.el6ost
errata-xmlrpc 2013-12-05 19:01:41 UTC Status MODIFIED ON_QA
Eoghan Glynn 2013-12-09 16:51:42 UTC Doc Text Cause: counter-intuitive rounding logic in the autoscaling calculation of server group growth/shrinkage for percentage adjustment.

Consequence: depending the choice of initial size and adjustment, the autoscaling group might never be scaled up regardless of the under-scaled alarm state.

Fix: intuitive rounding logic following the AWS model was implemented.

Result: regardless of the choice of initial size and adjustment, the autoscaling group will always be eligible for up- and down-scaling.
Eoghan Glynn 2013-12-11 16:52:42 UTC Status ON_QA VERIFIED
Don Domingo 2013-12-13 01:19:49 UTC CC ddomingo
Doc Text Cause: counter-intuitive rounding logic in the autoscaling calculation of server group growth/shrinkage for percentage adjustment.

Consequence: depending the choice of initial size and adjustment, the autoscaling group might never be scaled up regardless of the under-scaled alarm state.

Fix: intuitive rounding logic following the AWS model was implemented.

Result: regardless of the choice of initial size and adjustment, the autoscaling group will always be eligible for up- and down-scaling.
The Orchestration engine did not use proper rounding logic when using PercentChangeInCapacity to autoscale server group sizes. Specifically, autoscaling did not correctly round up if:

(MinSize x ScalingAdjustment / 100.00) < 1.0

As a result, certain combinations of MinSize and ScalingAdjustments could incorrectly prevent under-scaled groups from scaling up.

This fix adds the necessary rounding logic to PercentChangeInCapacity. As a result, PercentChangeInCapacity now correctly scales up as needed, regardless of the MinSize and ScalingAdjustment.
errata-xmlrpc 2013-12-19 17:39:15 UTC Status VERIFIED RELEASE_PENDING
errata-xmlrpc 2013-12-20 00:35:29 UTC Status RELEASE_PENDING CLOSED
Resolution --- ERRATA
Last Closed 2013-12-19 19:35:29 UTC
John Skeoch 2014-02-02 22:40:58 UTC CC srevivo

Back to bug 1030064