[nova][ptg] Summary: Using Forbidden Aggregates

Patil, Tushar Tushar.Patil at nttdata.com
Fri May 3 17:58:44 UTC 2019


>> - Spec has overall support, but a few open questions. Answer those, and
>> we're good to approve and move forward.
I have replied to all open questions and fix the nits.
I have accepted Tetsuro suggestion to add traits to the compute node resource provider in the nova placement sync_aggregates command if aggregates are configured with metadata with kye/value pair "trait:traits_name=required".

Request everyone to kindly review the updated specs.
https://review.opendev.org/#/c/609960/

Regards,
Tushar Patil

________________________________________
From: Eric Fried <openstack at fried.cc>
Sent: Friday, May 3, 2019 12:59:15 PM
To: OpenStack Discuss
Subject: [nova][ptg] Summary: Using Forbidden Aggregates

Spec: https://review.opendev.org/#/c/609960/

Summary:
- TL;DR: Allows you to say "You can't land on a host that does X unless
you specifically require X". Example: Keep my Windows-licensed hosts for
Windows instances.
- Exploit placement enablement for forbidden aggregates [1] in Nova
- Set (nova) aggregate metadata with a syntax similar/identical to that
of extra_specs for required traits (e.g. 'trait:CUSTOM_WINDOWS_ONLY':
'required')
- During scheduling, nova will discover all aggregates with metadata of
this form. For each:
  - Construct a list of the traits in the aggregate metadata
  - Subtract traits required by the server request's flavor+image.
  - If any traits from the aggregate remain, add this aggregate's UUID
(which corresponds to a placement aggregate) to the list of "forbidden
aggregates" for the GET /allocation_candidates request.

Agreements:
- The "discover all aggregates" bit has the potential to be slow, but is
better than the alternative, which was having the admin supply the same
information in a confusing conf syntax. And if performance becomes a
problem, we can deal with it later; this does not paint us into a corner.
- Spec has overall support, but a few open questions. Answer those, and
we're good to approve and move forward.

efried

[1]
https://docs.openstack.org/placement/latest/specs/train/approved/2005297-negative-aggregate-membership.html

Disclaimer: This email and any attachments are sent in strictest confidence for the sole use of the addressee and may contain legally privileged, confidential, and proprietary data. If you are not the intended recipient, please advise the sender by replying promptly to this email and then delete and destroy this email and any attachments without any further use, copying or forwarding.



More information about the openstack-discuss mailing list