make build_failure_weight_multiplier configurable

Bug #1892934 reported by Andrea Ieri
16
This bug affects 2 people
Affects Status Importance Assigned to Milestone
OpenStack Nova Cloud Controller Charm
Fix Released
Undecided
Andrea Ieri

Bug Description

The current charm behavior is to disable weighing down computes that experienced build failures. This was implemented as a fix for bug 1818239, after a lengthy discussion and with only a partial agreement.

In environments that more closely monitor load distribution throughout the cloud and keep track of failed computes, the previous behavior may actually be preferable, as it automates skipping badly configured or malfunctioning hypervisors, without requiring manual intervention (see in particular comment #21 from that discussion).

I propose making build_failure_weight_multiplier a configurable option, since it would allow us to support both use cases.

Andrea Ieri (aieri)
Changed in charm-nova-cloud-controller:
assignee: nobody → Andrea Ieri (aieri)
Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix proposed to charm-nova-cloud-controller (master)

Fix proposed to branch: master
Review: https://review.opendev.org/748310

Changed in charm-nova-cloud-controller:
status: New → In Progress
Revision history for this message
Andrea Ieri (aieri) wrote :

subscribed field-medium to mark a contribution needing to be merged

Revision history for this message
Andrea Ieri (aieri) wrote :

it would also be useful to add monitoring for demoted hypervisors, otherwise they may be forgotten until the cloud is so full that the scheduler has run out of "good" computes

Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix merged to charm-nova-cloud-controller (master)

Reviewed: https://review.opendev.org/748310
Committed: https://git.openstack.org/cgit/openstack/charm-nova-cloud-controller/commit/?id=1bf5a928325990ecb217b31137c3715bec91c6ff
Submitter: Zuul
Branch: master

commit 1bf5a928325990ecb217b31137c3715bec91c6ff
Author: Andrea Ieri <email address hidden>
Date: Tue Aug 25 18:39:30 2020 -0400

    Make disabling build_failure_weight_multiplier optional.

    This commit introduces a new charm option allowing operators to override
    the hardcoded 0.0 that disabled hypervisor demotion on build failures
    from pike onward.

    In certain environments it may be preferable to retain the upstream
    behavior of letting the scheduler work around malfunctional computes and
    favor instance building reliability at the cost of a potentially uneven
    load distribution.

    Change-Id: I2faa5ab8cd505a9d61a9fa26e1b08d16b0c795fb
    Closes-Bug: 1892934

Changed in charm-nova-cloud-controller:
status: In Progress → Fix Committed
Changed in charm-nova-cloud-controller:
milestone: none → 20.10
Changed in charm-nova-cloud-controller:
status: Fix Committed → Fix Released
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.