Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-04-26
14:33:00 mriedem dansmith: yes, i'm saying move the python condition into the sql query
14:33:09 openstackgerrit John Garbutt proposed openstack/nova master: Add description to policies in migrate_server.py https://review.openstack.org/459711
14:33:10 mriedem so we're pulling less stuff we don't care about over rpc
14:33:22 mriedem like with a bunch of joins and shit we don't care about
14:33:30 dansmith mriedem: but it's a specific list of instance uuids and we care about all of them because of ^
14:33:32 mriedem this is really just an existence check
14:33:37 dansmith it's not
14:33:50 dansmith because we need to delete allocations for instances that have been evacuated in our absence to another host
14:34:04 dansmith or instances that were locally deleted while we were gone
14:34:18 dansmith but not instances that are in flight (=None) or assigned to us while we were doing this (=$host)
14:34:24 mriedem i'll have to wait for the code change then to think about it
14:34:41 dansmith we have a call right now anyway
14:35:17 mriedem sure, i just want to make sure this is as fast as possible
14:35:22 mriedem expected_attrs=[] ftw
14:35:32 dansmith it will almost always be a no-op, but sure
14:48:54 bauzas mriedem: mmm, unless I misunderstood, you wanted to ask to use get_by_filters() by saying "not this host", right?
14:49:29 bauzas mriedem: but AFAIK you can use a filter like 'host'='foo', not 'host'= NOT 'foo'
14:49:40 mriedem host=None
14:49:47 mriedem we do that elsewhere in the code
14:50:11 bauzas mriedem: a-ha, I see
14:50:58 bauzas mriedem: but I also need to verify host=me
14:51:18 mriedem then i guess you can't filter in sql
14:51:36 bauzas yup, my point
14:51:39 bauzas I was confused
14:51:41 bauzas anyway
14:52:54 openstackgerrit John Garbutt proposed openstack/nova master: Add description to policies in migrate_server.py https://review.openstack.org/459711
14:54:03 mriedem bauzas: just make sure you pass expected_attrs=[] so we don't do a bunch of unnecessary joins
14:54:10 bauzas I know
14:57:11 dansmith you guys are optimizing the heck out of a list of <=1 :P
14:57:12 jaypipes dansmith, cdent: answered your questions on https://review.openstack.org/#/c/457782
14:58:26 mriedem lyarwood: can you hit some of these when you get a chance? https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:stable/newton
14:58:49 cdent thanks jaypipes
14:59:26 bauzas cdent: correct me if I'm wrong but...
14:59:47 bauzas cdent: is https://github.com/openstack/nova/blob/master/nova/scheduler/client/report.py#L871 returning a list of instance UUIDs ?
15:00:15 openstackgerrit Huan Xie proposed openstack/nova master: WIP: XenAPI use os-xenapi v2 in nova https://review.openstack.org/453493
15:00:42 cdent bauzas it's more than that, but does include instance uuids
15:01:03 bauzas cdent: it's keyed by that ? https://github.com/openstack/nova/blob/master/nova/scheduler/client/report.py#L880
15:01:14 mdbooth cdent jaypipes: Any reason we can't return structured objects from get_inventory(): https://review.openstack.org/#/c/457782/ ?
15:01:39 mdbooth dicts with grab bags of fields returned by drivers make me nervous
15:01:55 cdent bauzas: https://github.com/openstack/nova/blob/master/nova/api/openstack/placement/handlers/allocation.py#L138
15:02:00 lyarwood mriedem: ack, sorry did mean to get around to that earlier in the week.
15:02:21 bauzas cdent: cool then
15:02:21 jaypipes mdbooth: eventually, there will be a bit more structure there, when nested resource providers lands this patch: https://review.openstack.org/#/c/415920/
15:02:37 bauzas cdent: FWIW the gabbi tests for allocations are a bit long to read :)
15:02:42 cdent mdbooth: we're trying to kill off nova-side use of placement-used ovos, especially when rcp not involved
15:02:47 jaypipes mdbooth: but we didn't want to depend on versioned objects defined in nova/objects/resource_provider.py within the virt driver layer.
15:03:13 jaypipes bauzas: compared to...?
15:03:30 bauzas jaypipes: for knowing which allocations to delete
15:03:40 bauzas anyway, patch is close to be up
15:03:58 jaypipes bauzas: no, I mean the gabbits are hard to read, compared to what? :)
15:04:19 cdent bauzas: see line 51 of https://etherpad.openstack.org/p/placement-newton-leftovers
15:04:20 bauzas compared to API docs :)
15:04:23 mriedem there is a note in there that the get allocations gabbits can't sub the left side of the json path b/c of a bug
15:04:40 mriedem they are working on api-ref
15:04:50 cdent s/they/we/
15:04:57 cdent as in, all of us
15:05:13 mriedem us vs them
15:05:15 mriedem duh
15:05:31 cdent mriedem: are you the wolf or the dog today?
15:05:41 mriedem i meant, they as in 'people that aren't me'
15:05:42 mriedem :)
15:05:46 mriedem b/c i'm not being helpful
15:06:43 cdent the docs, although not yet published, are coming together nicely and look very nice and polished
15:08:21 cdent given sufficient time/energy/people it would be great to decompose the gabbi files into smaller pieces, been on the looking for volunteers todo list for months
15:10:41 openstackgerrit John Garbutt proposed openstack/nova master: Add description to policies in floating_ip files https://review.openstack.org/459791
15:12:11 efried mordred Just to be clear, for the endpoints-from-service-catalog stuff, you're suggesting changes in keystoneauth1 itself?
15:12:41 sean-k-mooney jaypipes: hi jay yes im familar with what at&t want to do we were talking about this at the PTG
15:12:47 openstackgerrit John Garbutt proposed openstack/nova master: Add description to policies in migrations.py https://review.openstack.org/459727
15:13:29 sean-k-mooney jaypipes: its been a while since i reviewd the code changes though so i need to see how its shaping up and what the apis will look like.
15:16:00 jaypipes sean-k-mooney: will you be in Boston?
15:16:19 sean-k-mooney i was not planning to be if i had a strong reason to go i could
15:16:56 jaypipes sean-k-mooney: heh, you know it's in <2 weeks right? :)
15:17:37 sean-k-mooney yep i know i do have a ticket for said event but others are going form my team just not me
15:18:04 sean-k-mooney i was not sure how much design was going to happen at the fourm now with the PTG
15:18:17 openstackgerrit Rafael Folco proposed openstack/nova master: Delete inventory has its own method DELETE https://review.openstack.org/460147
15:18:32 jaypipes sean-k-mooney: certainly less than the PTG. but if you were going to be there, I would propose meeting in person.
15:18:45 jaypipes sean-k-mooney: not a big deal though. we can always do a hangout.
15:18:57 openstackgerrit Ildiko Vancsa proposed openstack/nova master: WIP/POC Implement new attach Cinder flow https://review.openstack.org/330285
15:20:06 openstackgerrit Kaitlin Farr proposed openstack/nova master: Parse algorithm from cipher for ephemeral disk encryption https://review.openstack.org/446072
15:20:35 sean-k-mooney yep it would have been good to sync on things but ya im always up for a hangout's chat too
15:21:37 sean-k-mooney jaypipes: AT&T originally were proposing this as a nova change and i redirected it to neutron at the ptg.
15:22:03 jaypipes sean-k-mooney: k
15:22:29 mordred efried: I am - I think we should add a method to keystoneauth which registers options needed by "adapters" (which is keystoneauth term for a session mounted on a service endpoint)
15:22:54 mordred efried: so basically the parameters you need to create an adapter are the same parameters you need to find an endpoint from the service catalog
15:23:30 mordred and if we add them there, then we can just re-use that method ot register the right opts in all the places we want to let nova users put in all of the different filters
15:23:39 mordred (similar to how we use register_session_opts now)
15:24:04 efried mordred I dig it. Can you point me to something in the keystone source that'll give me a hint as to how an "adapter" differs from a "session" or an "auth"?
15:24:17 mriedem swebster: see my comment in https://review.openstack.org/#/c/453938/2/nova/network/neutronv2/api.py after the fact - can you push a small change to update that code comment block for the migration = None case?
15:24:26 mordred one could also make the argument that an load_adapter_from_opts would be a nice equivilent method ot keystoneauth that we could use in the places we want just a plain keystoneuath REST interface
15:24:28 openstackgerrit Gábor Antal proposed openstack/nova master: [WIP] Transform instance.live_migration_rollback notification https://review.openstack.org/402124
15:24:33 mordred efried: absolutely! one sec ...
15:25:09 swebster mriedem: no problem, will do
15:25:37 mordred efried: so - if you start from here: https://github.com/openstack/keystoneauth/blob/master/keystoneauth1/loading/__init__.py#L31-L32
15:25:46 efried mordred And next question: Where do I propose changes? review.openstack.org via https://git.openstack.org/openstack/keystoneauth ?
15:26:13 efried mordred Yuh, found that, and saw how it was being used by other groups (cinder, ironic).
15:26:18 mordred efried: you can see this: https://github.com/openstack/keystoneauth/blob/master/keystoneauth1/loading/session.py
15:26:37 mordred so I believe what would be wanted is a similar keystoneauth1/loading/adapter.py
15:27:25 mordred the actual adapter code is here: https://github.com/openstack/keystoneauth/blob/master/keystoneauth1/adapter.py - which lists the different adapter constructor params - I'm not sure all of those need to be conf options though
15:28:16 mordred efried: I'd say for your purposes, https://github.com/openstack/keystoneauth/blob/master/keystoneauth1/adapter.py#L32-L37 are the ones that want conf options
15:28:56 efried mordred Except service_name, which people thought was useless?

Earlier   Later