yahoo-eng-team team mailing list archive

[Sylvain Bauza <sbauza@xxxxxxx>]

That's a design discussion IMHO, not really a bug. In case you want to
have a better performance because of the DB queries, please use the
CachingScheduler instead of the FilterScheduler.

The CachingScheduler is good for not calling by every request the host
states.


** Changed in: nova
       Status: New => Opinion

** Tags added: scheduler

-- 
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/1568188

Title:
  Scheduler should respect the setting of
  scheduler_tracks_instance_changes

Status in OpenStack Compute (nova):
  Opinion

Bug description:
  The feature track instance change introduced in KILO has performance
  issue when handling larger env when hypervisor count > 300.

  The scheduler is still very slow even
  scheduler_tracks_instance_changes=False because every scheduling will
  invoke host_manager.get_all_host_states  and will invoke
  _add_instance_info for every host.

  Please adding one line to respect tracks_instance_change.
  class HostManager(BaseHostManager):
      ....
      def get_all_host_states(self, context):
              ....
              if self. tracks_instance_changes:
                  self._add_instance_info(context, compute, host_state)
              ....

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/1568188/+subscriptions