yahoo-eng-team team mailing list archive

[poovannan <2117406@xxxxxxxxxxxxxxxxxx>]

Public bug reported:

As per this changeset
https://github.com/openstack/nova/commit/953f1eef19139a6b7b47d56802f211008ca72e7c#diff-
ed9525d7ae319fd575249ed72daf634d27182e08fea7a4f740cb3164233612b7

We seem to have made change in resource_tracker.py. Specifically in _update_to_placement method, we are now calling 
allocs = self.reportclient.get_allocations_for_provider_tree(
            context, nodename)
all the time. If there are more virtual machines in the openstack environment, then this periodic call which gets executed makes GET allocations call for each VM. i.e Lot of GET calls are made to placement. 
Earlier in openstack yoga, it was not the case as only if reshape is needed, call is made.

As per comments in commit, 
So the code
is not prepared to do a single, one shot, reshape at startup, but
instead to do a continuous healing of the allocations. We can remove
this continuous healing after the PCI prefilter will be made mandatory
in a future release.

Not sure when we plan to remove this. But can this be reverted if the
PCI prefilter is made mandatory ?

In our env (openstack caracal) : 
grep "GET /allocations/" placement-api.log | wc -l
153914

This count is too high compared to openstack yoga.
Current branch : https://github.com/openstack/nova/blob/master/nova/compute/resource_tracker.py also has the same change. It would be good if reshape can be optimal like earlier release.

** Affects: nova
     Importance: Undecided
         Status: New

-- 
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/2117406

Title:
  More placement allocations call in caracal

Status in OpenStack Compute (nova):
  New

Bug description:
  As per this changeset
  https://github.com/openstack/nova/commit/953f1eef19139a6b7b47d56802f211008ca72e7c#diff-
  ed9525d7ae319fd575249ed72daf634d27182e08fea7a4f740cb3164233612b7

  We seem to have made change in resource_tracker.py. Specifically in _update_to_placement method, we are now calling 
  allocs = self.reportclient.get_allocations_for_provider_tree(
              context, nodename)
  all the time. If there are more virtual machines in the openstack environment, then this periodic call which gets executed makes GET allocations call for each VM. i.e Lot of GET calls are made to placement. 
  Earlier in openstack yoga, it was not the case as only if reshape is needed, call is made.

  As per comments in commit, 
  So the code
  is not prepared to do a single, one shot, reshape at startup, but
  instead to do a continuous healing of the allocations. We can remove
  this continuous healing after the PCI prefilter will be made mandatory
  in a future release.

  Not sure when we plan to remove this. But can this be reverted if the
  PCI prefilter is made mandatory ?

  In our env (openstack caracal) : 
  grep "GET /allocations/" placement-api.log | wc -l
  153914

  This count is too high compared to openstack yoga.
  Current branch : https://github.com/openstack/nova/blob/master/nova/compute/resource_tracker.py also has the same change. It would be good if reshape can be optimal like earlier release.

To manage notifications about this bug go to:
https://bugs.launchpad.net/nova/+bug/2117406/+subscriptions