Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2019-04-16
16:51:43 efried dansmith: I mean allow reshapes to happen when they need to happen, rather than restricting them to upgrade boundaries. That's what bauzas is talking about as well.
16:52:05 efried Yes, dansmith and bauzas we still need to fail the reshape if it entails moving allocations in an impossible way.
16:52:05 dansmith efried: right I know, but I think there's subtlety here
16:52:19 bauzas efried: to be fair, VGPU reshapes are done on compute startup already, not upgrade: )
16:52:40 bauzas of course, it will in theory run once, after upgrading
16:52:49 dansmith efried: we have to maintain config compatibility, but if if we don't have information in the N-1 config to do the reshape, then we have to be able to punt the reshape (triggered by an upgrade) until after the config is updated
16:52:53 efried like if you suddenly specify your PCPU pinset to be empty, but have instances running with dedicated CPUs, that's a fail.
16:53:05 dansmith and, I agree that if you have to change how many pcpus are dedicated, we have to reshape again
16:53:16 efried I think we're on the same page
16:53:20 bauzas efried: that's why I proposed to check the allocations and inventories *before* providing the reshape
16:53:51 bauzas if the operator changes the config, but placement says "sorry but you can't", then the compute will fail to restart
16:54:22 dansmith efried: I think the thing I don't want, which I expressed as "only at upgrade time" is something like we reshape every time we restart compute because we decide we can arrange things better, or some state in the db has changed, but reshape due to a config/structural change makes sense
16:54:29 bauzas if the operator changes the config, and placement resources are okay, then the driver returns a ReshapeNeeded
16:54:35 efried dansmith: wfm
16:54:39 bauzas and then the new inventories and allocations
16:54:57 bauzas okay, so dansmith, efried and I are on the same page
16:55:25 bauzas there is one last concern from stephenfin about the config knob for the PCPU request
16:55:44 efried So e.g. in the PCPU spec, we're inferring the counts and pinsets of VCPUs vs PCPUs based on existing conf options.
16:55:49 bauzas but I think it's okay to make the request transformation to be "config-driven"
16:56:15 efried So the operator needs to change to the new config in such a way that it *exactly* matches what we inferred, right?
16:56:27 bauzas so, the operator would basically tell when he's okay to count PCPUs (ie. probably after the whole nodes config change)
16:56:59 efried Otherwise we don't just need a reshape (move allocations) - we would also possibly need to re-pin guests to different physical processors and such.
16:57:06 bauzas efried: no, I'm saying that existing config will report VCPUs anyway
16:57:20 bauzas (including options that were asking for pinned cpu)s
16:57:41 bauzas efried: only new config option (explicitely cpu_dedicated_set) will trigger a reshape
16:57:55 efried yes, I get that bauzas, what I'm saying is, we're going to *infer* VCPU/PCPU counts and pinsets based on legacy conf options; but then the operator wants to cut over to using the new conf options.
16:58:24 bauzas no, I don't want us to infer VCPU and PCPU based on those options because they are errorprone
16:58:28 bauzas efried: ^
16:58:38 bauzas those options being the legacy ones
16:58:49 efried oh, that's the basis for the PCPU spec as written at PS24 anyway. Haven't checked since then...
16:59:39 bauzas that's exactly why I'm saying "don't touch anything until operator explicitely says 'I want cpu_dedicated_set')
16:59:53 bauzas old world = VCPU
17:00:03 bauzas new world = VCPU and PCPU
17:00:30 bauzas for the request, trigger the request option when you consider having enough hosts to sustain PCPU requests
17:00:59 bauzas anyway, I need to bail out
17:01:24 bauzas kids aren't in town, and I promised some evening to my spouse
17:01:25 efried I think I see. Did you comment accordingly on the spec?
17:01:33 bauzas efried: I think so
17:01:36 efried okay.
17:03:01 sean-k-mooney efried: yes the cpu spec currently does infer but i agree with bauzas that we should not
17:04:02 sean-k-mooney efried: i think not supporting inference form the old config values would remove much of the upgrade impact or atleast help us too.
17:05:04 efried I'm good with that. Let's get it reflected in the spec
17:05:33 bauzas I just provided a comment trying to summarize my thoughts
17:05:45 bauzas this said, calling it a day
17:05:50 sean-k-mooney o/
17:06:23 efried imacdonn: I agree with your assessment in https://bugs.launchpad.net/nova/+bug/1824435
17:06:24 openstack Launchpad bug 1824435 in OpenStack Compute (nova) stein "fill_virtual_interface_list migration fails on second attempt" [High,Triaged]
17:06:32 efried (I wanted to say that, but not pollute the bug with it)
17:06:52 mriedem dansmith: you want to check my rolling upgrade validation logic that i dumped in gibi's spec https://review.openstack.org/#/c/652608/4/specs/train/approved/server-move-operations-with-ports-having-resource-request.rst@190 ?
17:10:22 dansmith mriedem: I probably have to read the whole spec to make sense of that huh?
17:11:04 imacdonn efried, which one? i.e. do you think we need to address (2) or would fixing (1) obviate that ?
17:11:37 efried imacdonn: fixing (1) would obviate. But that assumes we can do so.
17:12:05 efried imacdonn: IMO we should fix (1) and change (2) to raise an explicit exception to "guarantee" it.
17:12:05 imacdonn efried, right ... so now I'm trying to understand why the row is being created in the first place
17:12:09 efried ++
17:12:48 efried imacdonn: Is it possible for the rows to differ in any material way?
17:13:01 efried (i.e. a way that makes a difference to the outcome)
17:13:05 mriedem dansmith: not really, it's just the usual "how could this fail during an upgrade"
17:13:18 mriedem dansmith: he needs to pass new parameters to compute rpc api methods,
17:13:30 dansmith well, I read it and seemed like I needed to understand more, so I'm reading the wholething now
17:13:33 mriedem which could be (1) stein computes that don't handle those or (2) rpc pinned so we pop those parameters
17:13:36 mriedem ok
17:13:47 imacdonn efried, I'm assuming that _security_group_get_by_names() is used elsewhere (or at least intended to be reusable), so probably should consider all possible use-cases, if we tackle that one
17:14:45 efried used in two places
17:15:19 mriedem imacdonn: efried: honestly i'm not sure how much relevance that code even has anymore if you're using neutron
17:15:36 imacdonn mriedem, I was wondering about that
17:16:31 mriedem i think it at least means if you're using neutron, every project that ever created an instance in nova has a 'default' security_groups table record that is never used or cleaned up
17:16:33 mriedem vestigial
17:17:12 imacdonn that seems plausible
17:22:48 dansmith mriedem: see if my words help at all
17:26:47 openstackgerrit Stephen Finucane proposed openstack/nova master: Remove '/os-cells' REST APIs https://review.openstack.org/651291
17:26:47 openstackgerrit Stephen Finucane proposed openstack/nova master: Stop handling cells v1 in '/os-hypervisors' API https://review.openstack.org/651292
17:26:48 openstackgerrit Stephen Finucane proposed openstack/nova master: Stop handling cells v1 in '/os-servers' API https://review.openstack.org/651293
17:26:48 openstackgerrit Stephen Finucane proposed openstack/nova master: Remove 'nova-manage cell' commands https://review.openstack.org/651294
17:26:49 openstackgerrit Stephen Finucane proposed openstack/nova master: Stop handling cells v1 for console authentication https://review.openstack.org/651295
17:26:49 openstackgerrit Stephen Finucane proposed openstack/nova master: Remove old-style cell v1 instance listing https://review.openstack.org/651296
17:26:50 openstackgerrit Stephen Finucane proposed openstack/nova master: Remove 'bdm_(update_or_create|destroy)_at_top' https://review.openstack.org/651297
17:26:50 openstackgerrit Stephen Finucane proposed openstack/nova master: Remove 'instance_fault_create_at_top' https://review.openstack.org/651298
17:26:51 openstackgerrit Stephen Finucane proposed openstack/nova master: Remove 'instance_info_cache_update_at_top' https://review.openstack.org/651299
17:26:51 openstackgerrit Stephen Finucane proposed openstack/nova master: Remove 'get_keypair_at_top' https://review.openstack.org/651300
17:26:52 openstackgerrit Stephen Finucane proposed openstack/nova master: Remove 'instance_update_at_top', 'instance_destroy_at_top' https://review.openstack.org/651301
17:26:52 openstackgerrit Stephen Finucane proposed openstack/nova master: Remove 'instance_update_from_api' https://review.openstack.org/651302
17:26:53 openstackgerrit Stephen Finucane proposed openstack/nova master: Stop handling 'update_cells' on 'BandwidthUsage.create' https://review.openstack.org/651303
17:26:53 openstackgerrit Stephen Finucane proposed openstack/nova master: Stop handling cells v1 for instance naming https://review.openstack.org/651304
17:26:54 openstackgerrit Stephen Finucane proposed openstack/nova master: Remove cells code https://review.openstack.org/651306
17:26:54 openstackgerrit Stephen Finucane proposed openstack/nova master: Stop handling 'InstanceUnknownCell' exception https://review.openstack.org/651307
17:26:55 openstackgerrit Stephen Finucane proposed openstack/nova master: Remove unnecessary wrapper https://review.openstack.org/651308
17:26:55 openstackgerrit Stephen Finucane proposed openstack/nova master: db: Remove cell APIs https://review.openstack.org/651309
17:27:08 mriedem dansmith: replied but yeah i think so
17:27:59 dansmith cool
17:28:42 sean-k-mooney mriedem: so related but sperate form gibis spec should we jsut use teh livemigration multiple port bindign workflow for all move operations
17:29:04 sean-k-mooney mriedem: i mentioned it in gibi's spec but it would simplfy things to have a singel common code path
17:29:53 mriedem given the issues NewBruce is hitting idk
17:30:36 mriedem moving all move operations to that model would be a big change i think, and likely not something we should block gibi's spec on
17:30:53 sean-k-mooney oh im not suggesting we should block on it
17:31:11 sean-k-mooney its jsut some/alot of the compleix would be reduced
17:31:37 sean-k-mooney but NewBruce issue is concering ill grant that
17:35:28 francoisp_ alex_xu, hi, would you have time to look at https://review.openstack.org/#/c/648123/6 - thanks!
18:05:11 openstackgerrit Matt Riedemann proposed openstack/nova master: Do not create default security group during instance create if using Neutron https://review.openstack.org/653065
18:05:12 mriedem imacdonn: let's see what blows up ^

Earlier   Later