Back to bug 1306263

Who When What Removed Added
Red Hat Bugzilla Rules Engine 2016-02-10 12:39:46 UTC Flags testing_ack? planning_ack?
Martin Sivák 2016-02-10 12:40:44 UTC Blocks 1207255
Assignee nobody msivak
Doron Fediuck 2016-02-21 08:36:55 UTC CC dfediuck
Target Milestone --- ovirt-4.0.0
Flags ovirt-4.0.0?
Aharon Canan 2016-03-13 15:37:59 UTC QA Contact acanan mavital
Sandro Bonazzola 2016-05-02 09:58:13 UTC Target Milestone ovirt-4.0.0-alpha ovirt-4.0.0-beta
Yaniv Lavi 2016-05-23 13:19:09 UTC Target Milestone ovirt-4.0.0-beta ovirt-4.0.0-rc
Yaniv Kaul 2016-06-15 07:32:50 UTC Target Milestone ovirt-4.0.0-rc ovirt-4.0.2
Flags ovirt-4.0.0? ovirt-4.0.z?
Martin Sivák 2016-07-13 09:42:12 UTC Status NEW ASSIGNED
Target Milestone ovirt-4.0.2 ovirt-4.1.0-alpha
Martin Sivák 2016-07-13 09:42:52 UTC Flags ovirt-4.0.z? ovirt-4.1.0?
Andrej Krejcir 2016-10-18 14:14:46 UTC CC akrejcir
Martin Sivák 2016-11-07 12:34:36 UTC Blocks 1392393
Red Hat Bugzilla Rules Engine 2016-12-01 17:12:07 UTC Link ID oVirt gerrit 67707
Status ASSIGNED POST
Flags planning_ack? planning_ack+
meital avital 2016-12-03 07:56:33 UTC QA Contact mavital alukiano
Flags testing_ack? testing_ack+
Red Hat Bugzilla Rules Engine 2016-12-03 07:56:41 UTC Flags ovirt-4.1? ovirt-4.1+
Martin Sivák 2016-12-05 14:34:12 UTC Status POST MODIFIED
URL http://www.ovirt.org/develop/release-management/features/sla/scheduling-weight-normalization/
Doc Text Feature:

The weighting step of VM scheduling changed and the best host is now selected using weighted rank algorithm instead of a pure sum of weights.

Separate Rank is computed for policy unit and host and represents the number of hosts that were worse according to the policy unit.

The weight multiplier is then used to multiply the ranks for the given policy unit.

The host with greatest sum of ranks wins.

For details please see: http://www.ovirt.org/develop/release-management/features/sla/scheduling-weight-normalization/

Reason:

The currently used weight policy units do not use any common result value range. Each unit reports numbers as needed and this causes issues with user configured preferences (policy unit multiplier), because for example memory (very high numbers) always wins over CPU (0-100).

Result:

The impact of the policy unit multiplier (from the scheduling policy configuration) is much more predictable. However users that are using it should check their configuration for sanity.
Doc Type Bug Fix Enhancement
Martin Sivák 2016-12-14 11:20:03 UTC Target Milestone ovirt-4.1.0-alpha ovirt-4.1.0-beta
Martin Sivák 2016-12-14 11:38:56 UTC Blocks 1404660
Gil Klein 2016-12-28 21:40:59 UTC Status MODIFIED ON_QA
CC gklein
Artyom 2017-01-03 16:19:39 UTC Keywords Triaged
Status ON_QA VERIFIED
Sandro Bonazzola 2017-01-23 10:05:24 UTC Target Release --- 4.1.0.2
Byron Gravenorst 2017-01-24 01:00:08 UTC CC bgraveno, msivak
Doc Text Feature:

The weighting step of VM scheduling changed and the best host is now selected using weighted rank algorithm instead of a pure sum of weights.

Separate Rank is computed for policy unit and host and represents the number of hosts that were worse according to the policy unit.

The weight multiplier is then used to multiply the ranks for the given policy unit.

The host with greatest sum of ranks wins.

For details please see: http://www.ovirt.org/develop/release-management/features/sla/scheduling-weight-normalization/

Reason:

The currently used weight policy units do not use any common result value range. Each unit reports numbers as needed and this causes issues with user configured preferences (policy unit multiplier), because for example memory (very high numbers) always wins over CPU (0-100).

Result:

The impact of the policy unit multiplier (from the scheduling policy configuration) is much more predictable. However users that are using it should check their configuration for sanity.
The weighting for virtual machine scheduling has been updated. The best host for the virtual machine is now selected using a weighted rank algorithm instead of the pure sum of weights. A separate rank is calculated for the policy unit and host, and the weight multiplier is then used to multiply the ranks for the given policy unit. The host with the highest number is selected.

The reason for the change is that current weight policy units do not use a common result value range. Each unit reports numbers as needed, and this causes issues with user configured preferences. For example, memory (which has high numbers) always wins over CPU (0-100).

This update ensures that the impact of the policy unit multiplier for the scheduling policy configuration is more predictable. However, users that are using it should check their configuration for sanity when upgrading.
Flags needinfo?(msivak)
Martin Sivák 2017-01-24 09:44:12 UTC Flags needinfo?(msivak)
Sandro Bonazzola 2017-02-01 14:45:15 UTC Status VERIFIED CLOSED
Resolution --- CURRENTRELEASE
Last Closed 2017-02-01 09:45:15 UTC
Bronce McClain 2017-02-08 21:09:51 UTC CC bmcclain
Flags priority_rfe_tracking+
Byron Gravenorst 2017-03-28 05:34:54 UTC Doc Text The weighting for virtual machine scheduling has been updated. The best host for the virtual machine is now selected using a weighted rank algorithm instead of the pure sum of weights. A separate rank is calculated for the policy unit and host, and the weight multiplier is then used to multiply the ranks for the given policy unit. The host with the highest number is selected.

The reason for the change is that current weight policy units do not use a common result value range. Each unit reports numbers as needed, and this causes issues with user configured preferences. For example, memory (which has high numbers) always wins over CPU (0-100).

This update ensures that the impact of the policy unit multiplier for the scheduling policy configuration is more predictable. However, users that are using it should check their configuration for sanity when upgrading.
The weighting for virtual machine scheduling has been updated. The best host for the virtual machine is now selected using a weighted rank algorithm instead of the pure sum of weights. A rank is calculated for the policy unit and host, and the weight multiplier is then used to multiply the ranks for the given policy unit. The host with the highest number is selected.

The reason for the change is that current weight policy units do not use a common result value range. Each unit reports numbers as needed, and this causes issues with user configured preferences. For example, memory (which has high numbers) always wins over CPU (0-100).

This update ensures that the impact of the policy unit multiplier for the scheduling policy configuration is more predictable. However, users that are using it should check their configuration for sanity when upgrading.

Back to bug 1306263