[openstack-dev] [nova] Next steps for resource providers work

Chris Dent cdent+os at anticdent.org
Wed Aug 31 17:30:16 UTC 2016


> On 08/29/2016 12:40 PM, Matt Riedemann wrote:
>> I've been out for a week and not very involved in the resource providers
>> work, but after talking about the various changes up in the air at the
>> moment a bunch of us thought it would be helpful to lay out next steps
>> for the work we want to get done this week.

There was another hangout today where we caught up on where we are.
Some notes were added to the etherpad
https://etherpad.openstack.org/p/placement-next

There is code either merged or pending merge that allows the
resource tracker to ensure that resource providers exist and have
the correct inventory.

The major concern and blocker at this point is setting and deleting
allocations, for which the assistance of Jay is required. Some
details follow with a summary of Jay's todos at the bottom.

There are two patches, starting at
https://review.openstack.org/#/c/363209/

The first is a hack to get the object side handling for
AllocationList.create_all and delete_all. As noted in the comments
there we're not sure about the atomicity in create_all and need Jay
to determine if what's there can be made to work, or as suggested we
need a mondo SQL thing to get it right. If the latter, we need Jay
to write it :)

I'm going to carry on with those patches now and try to add some
generation handling back in to protect against inventory changing
out from under us while making allocations, but I'm not confident
of getting it anything more that possible adequate and great would
be better.

During that I'm also going to try to adjust things so that we can
update an existing allocation, not just create them, as we've
determined that's required. set_all, not create_all, basically.

The other missing piece is the client side of setting and deleting
allocations, from the resource tracker. We'd like Jay to start this
too or if we're all lucky maybe it is started already?

And finally there's a question we didn't know how to answer: What
will the process be for healing instances that already exist before
the placement service is started, and thus have no allocations?

So to summarize Jay's to do list (please and thank you very much):

* Look at https://review.openstack.org/#/c/363209/ and decide if it
   is good enough to get rolling or needs to be completely altered.
* If the latter, alter it.
* Write the allocation client.
* Consult on healing instance allocations.

Meanwhile several people are involved in related clean up patches in
both nova and devstack to smooth off rough edges while we pushed a
lot of code.

Thanks to everyone today for pushing so hard. We look pretty close to
getting the must haves happening.

-- 
Chris Dent               ┬─┬ノ( º _ ºノ)        https://anticdent.org/
freenode: cdent                                         tw: @anticdent


More information about the OpenStack-dev mailing list