yahoo-eng-team team mailing list archive

[OpenStack Infra <1513654@xxxxxxxxxxxxxxxxxx>]

Reviewed:  https://review.openstack.org/261108
Committed: https://git.openstack.org/cgit/openstack/nova/commit/?id=13c1301f41edee6559220f2dfa92211714925f8d
Submitter: Jenkins
Branch:    master

commit 13c1301f41edee6559220f2dfa92211714925f8d
Author: andrewbogott <abogott@xxxxxxxxxxxxx>
Date:   Mon Dec 21 22:04:40 2015 -0600

    Add a disk space weight-based scheduler.
    
    By default this is turned on and has equal weight to the
    ram weigher.
    
    DocImpact: Added the disk_weight_multiplier config option
    Closes-Bug: 1513654
    Change-Id: I29ce73122ad1860081b64b75646a297dfbb8d292


** Changed in: nova
       Status: In Progress => Fix Released

-- 
You received this bug notification because you are a member of Yahoo!
Engineering Team, which is subscribed to OpenStack Compute (nova).
https://bugs.launchpad.net/bugs/1513654

Title:
  scheduler:  disk_filter permits scheduling on full drives

Status in OpenStack Compute (nova):
  Fix Released

Bug description:
  I use qcow images and have disk_allocation_ratio == 2.1 to allow large
  amounts of overcommitting of disk space.  To quote the nova config
  reference:

  >  If the value is set to >1, we recommend keeping track of the free disk space, as the value approaching 0 may result in the incorrect 
  >     functioning of instances using it at the moment.

  Good advice, but 'keeping track' can be a bit impractical at times.  I
  just now had the scheduler drop a large sized instance onto a server
  with a 98% full drive since the behavior of disk_allocation_ratio
  intentionally ignores the actual free space on the drive.

  I propose that we add an additional config setting to the disk
  scheduler so that I can overschedule but can /also/ request that the
  scheduler stop piling things onto an already groaning server.

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1513654/+subscriptions