Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-08-28
19:27:58 artom Remove the top-level 'allocation' and just use a ListOfObjectsField?
19:28:07 mriedem artom: +1
19:28:22 artom I still think it depends on what you'll be doing with that
19:28:26 edleafe So the object will have to have a method for creating the POST body
19:28:36 artom As in, if the provider uuids are important at all, it could be a dict keyed on those
19:28:55 edleafe artom: this should be opaque. It is the body needed to claim/unclaim a selected host and its resources
19:29:20 mriedem creating an actual allocation request is pretty trivial https://review.openstack.org/#/c/496031/3/nova/conductor/tasks/live_migrate.py@122
19:29:22 artom Completely opaque? Just a string then?
19:29:34 mriedem we're doing that when bypassing the scheduler and forcing a host for live migration ^
19:29:35 cdent edleafe: should we fix this first: https://bugs.launchpad.net/nova/+bug/1708204
19:29:36 openstack Launchpad bug 1708204 in OpenStack Compute (nova) "placement allocation representation asymetric on PUT and GET" [Wishlist,Confirmed]
19:29:39 edleafe sure, we can chop this up any way that fits ovo, but then it has be able to put back together
19:30:18 edleafe mriedem: that won't be as easy with shared/nested providers
19:30:41 mriedem yeah i understand
19:31:00 mriedem and per the todo in that patch, i'd like to still call the scheduler even if we're forcing the host during live migration
19:31:06 edleafe the whole point of returning these allocation dicts was to handle those more complex situations
19:31:46 mriedem we already have an AllocationCandidate object today right?
19:32:12 mriedem sorry, it's called AllocationRequest
19:32:53 mriedem which has a list of AllocationRequestResource
19:33:09 edleafe mriedem: I guess I'll try following that pattern
19:33:20 mriedem why not just use the same exact objects?
19:33:30 mriedem if that's exactly what you want for the structure?
19:33:36 edleafe it's not
19:34:05 mriedem ok
19:34:11 edleafe I need a thing that is a hostname, and the blob needed to claim/unclaim the resources for the particular request
19:34:49 mriedem "(2:22:02 PM) cdent: edleafe: so you’re idea at this point is an object that is effectively; ‘target’: String, ‘allocation’: Dict"
19:35:04 edleafe mriedem: yep
19:35:08 mriedem and then return a list of those
19:35:19 mriedem AlternativeHost, and AlternativeHostList
19:35:21 mriedem or something
19:35:35 edleafe we can bikeshed on the name :)
19:35:56 mriedem and you're wondering if you need to break down the 'allocation' field into a more specific type?
19:36:05 edleafe what I was asking is: what ovo field type do I use for the allocation part of that?
19:36:21 edleafe I don't want to break it down if at all possible
19:36:27 edleafe It's a unit
19:36:39 edleafe (for a particular claim)
19:37:01 artom edleafe, so a single allocation (ie, resource_provider and resources)
19:37:02 artom ?
19:37:24 edleafe artom: well, it would be a set of allocations, but yes
19:37:28 mriedem edleafe: then it's just Dict
19:37:53 mriedem edleafe: and this is an allocation, yes?
19:37:54 mriedem { "resource_provider": { "uuid": "9cf544dd-f0d7-4152-a9b8-02a65804df09" }, "resources": { "VCPU": 2, "MEMORY_MB": 8096 } }
19:37:59 edleafe mriedem: even with dansmith's note at https://github.com/openstack/nova/blob/master/nova/objects/fields.py#L72 ?
19:38:12 mriedem edleafe: yes, as i said, that's about using the nova.fields vs ovo fields
19:38:24 mriedem versioned objects started in nova,
19:38:33 mriedem that note is saying, stuff is moving to ovo, so use the fields from there directly
19:38:56 edleafe mriedem: ok, then that note confused me
19:38:56 mriedem https://github.com/openstack/oslo.versionedobjects/blob/master/oslo_versionedobjects/fields.py#L657
19:39:23 edleafe everyone back to work!
19:39:51 mriedem update the note with "Use these directly from oslo.versionedobjects for new code." or something if that clarifies it
19:41:00 edleafe mriedem: will do
19:42:06 sdague mriedem: https://review.openstack.org/#/c/498477 and https://review.openstack.org/#/c/498478 are redone with the git sha
19:42:40 mriedem sdague: ok, but i'm going to hold off on stable/pike stuff until we release
19:42:51 sdague mriedem: that's fine
19:43:02 sdague it will still have real test results on it
19:43:19 openstackgerrit Steve Noyes proposed openstack/nova master: update live migration to use v3 cinder api https://review.openstack.org/463987
19:43:24 sdague I was less concerned for the release, but just more for "these things should be in there sometime"
19:46:08 openstackgerrit Ed Leafe proposed openstack/nova master: Clarify the field usage guidelines https://review.openstack.org/498581
19:46:10 edleafe mriedem: quick note clarification ^^
20:24:47 openstackgerrit Dan Smith proposed openstack/nova master: Add placeholder migrations for Pike backports https://review.openstack.org/496932
20:24:47 openstackgerrit Dan Smith proposed openstack/nova master: Add uuid to migration object and migrate-on-load https://review.openstack.org/496934
20:24:48 openstackgerrit Dan Smith proposed openstack/nova master: Add uuid online migration for migrations https://review.openstack.org/498589
20:24:48 openstackgerrit Dan Smith proposed openstack/nova master: Add uuid to migration table https://review.openstack.org/496933
20:25:02 dansmith mriedem: that placeholder one failed out of the gate if you want to hit it again ^
20:25:40 mnaser (sorry, trying to get some help from another project that might have ran into this: i've discovered a performance issue in sahara and identified it to sqlalchemy lazy='joined' in the relationships which results in a giant set of joins)
20:25:52 mnaser i noticed the nova code had none of that. was that a decision that was taken or just.. no one ever set that to anything
20:26:13 mnaser i figure nova is probably one of the longer codebases out here that might have ran into these perofmance regressions so i'd ask around
20:26:40 dansmith mnaser: we generally join things explicitly when we want them
20:27:06 dansmith intentionally.
20:28:14 mnaser dansmith makes perfet sense. looking at the git blame.. looks like it was just magically changed with no explanation. ill do my checks
20:28:16 mnaser thank you for that info
20:49:28 openstackgerrit Matt Riedemann proposed openstack/nova master: Refactor ServerMovingTests for non-move tests https://review.openstack.org/498596
20:50:29 mriedem dansmith: ^ as you asked for
20:50:38 dansmith ah thanks
20:52:12 openstackgerrit Michael Still proposed openstack/nova master: Move lvm handling to privsep. https://review.openstack.org/495516
20:52:13 openstackgerrit Michael Still proposed openstack/nova master: Move shred to privsep. https://review.openstack.org/495537
20:52:13 openstackgerrit Michael Still proposed openstack/nova master: Move xend existence probes to privsep. https://review.openstack.org/495538
20:52:14 openstackgerrit Michael Still proposed openstack/nova master: WIP / Aspirational: we don't need rootwrap any more. https://review.openstack.org/495542
20:52:14 openstackgerrit Michael Still proposed openstack/nova master: Cleanup mount / umount and associated rmdir calls https://review.openstack.org/494423
20:52:15 openstackgerrit Michael Still proposed openstack/nova master: Move loopback setup and removal to privsep. https://review.openstack.org/495664
20:52:16 openstackgerrit Michael Still proposed openstack/nova master: Don't shell out to mkdir, use ensure_tree() https://review.openstack.org/492326
20:52:16 openstackgerrit Michael Still proposed openstack/nova master: Move libvirt usages of chown to privsep. https://review.openstack.org/471972
20:52:17 openstackgerrit Michael Still proposed openstack/nova master: Move ploop commands to privsep. https://review.openstack.org/492325
20:52:17 openstackgerrit Michael Still proposed openstack/nova master: Move the idmapshift binary into privsep. https://review.openstack.org/495541
20:52:18 openstackgerrit Michael Still proposed openstack/nova master: Read from console ptys using privsep. https://review.openstack.org/489486
20:52:21 openstackgerrit Michael Still proposed openstack/nova master: Avoid chowning console logs in libvirt https://review.openstack.org/472229
20:52:22 openstackgerrit Michael Still proposed openstack/nova master: First attempt at adding a privsep user to nova itself. https://review.openstack.org/459166
20:52:23 openstackgerrit Michael Still proposed openstack/nova master: Move execs of touch to privsep. https://review.openstack.org/489190
20:52:24 openstackgerrit Michael Still proposed openstack/nova master: Move libvirts dmcrypt support to privsep. https://review.openstack.org/490737
20:52:24 openstackgerrit Michael Still proposed openstack/nova master: Replace dd with shred for zeroing lvm volumes. https://review.openstack.org/495532
20:52:26 openstackgerrit Michael Still proposed openstack/nova master: Move execs of tee to privsep. https://review.openstack.org/489438
20:53:10 mikal sdague: that's me rebasing that series, thanks for the review on the first one
20:53:53 mikal sdague: https://review.openstack.org/#/c/495532/ is a similarly trivial head-of-series refactor if you're bored
20:54:23 mikal sdague: well, "refactor" might be strong. "removal of historical silliness" might be more accurate.
20:55:19 sdague sure, it's about end of day here, so I can look tomorrow
20:55:27 edleafe mikal: I prefer "hysterical silliness", myself
21:00:37 mriedem mikal: your test removal in the bottom change cuts a bit too deep
21:03:48 mriedem wow
21:03:49 mriedem "Took 269.99 seconds to spawn the instance on the hypervisor."
21:04:08 mriedem clarkb: ^

Earlier   Later