Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2018-05-31
18:58:26 esberglu mriedem: ack. Thanks for the review!
19:01:00 mriedem melwitt: with ^ the powervm series can be taken out of the runway
19:01:32 melwitt awesome
19:02:24 melwitt so next is going to be cert validation or cpu features as traits ... cert validation recently landed the spec amendment so should be good to go. thoughts?
19:06:22 cfriesen eandersson: you asked earlier about CPU pinning. In upstream the only way to see it is in the nova-compute logs or by looking at the DB.
19:07:19 cfriesen eandersson: internally we found it useful information and added a way to get it via the API, but never created a spec/blueprint for it.
19:20:36 mriedem melwitt: yes to cert validation
19:20:55 melwitt thanks
19:21:23 mriedem cfriesen: to which API?
19:24:59 dansmith I would like to see cert validation get another go and hopefully get finished
19:26:19 melwitt me too. thanks for the input, will move that into the runway next
19:29:21 melwitt mriedem: I'm working on getting the osc queens + novaclient queens stuff working with these patches https://review.openstack.org/#/q/topic:openstackclient-novaclient+(status:open+OR+status:merged) but from what I understand, nothing in the gate will verify those go together
19:30:29 melwitt I was thinking I might need to do some DNM patches to devstack to pull osc from git stable/queens and then a DNM revert of this https://review.openstack.org/538541 in grenade with depends-on to do it. would that work? is there a better way?
19:34:13 openstackgerrit Dan Smith proposed openstack/nova master: Use oslo.messaging per-call monitoring https://review.openstack.org/566696
19:38:35 mriedem melwitt: your question is breaking my brain
19:39:08 mriedem you want to test osc 3.14.2 with novaclient 10.1.0?
19:39:29 mriedem in queens, novaclient 10.x broke grenade and horizon, which got fixed in queens,
19:39:34 melwitt sorry ... that was too low-level a way to ask. yes, but osc 3.14.2 is not released yet, I was hoping to be able to verify it solves the problem before landing it if possible
19:39:40 mriedem the only remaining things that i knew of from the 10.x fallout was osc
19:40:22 melwitt yeah, that's my understanding too. I think it will work but realized I don't think anything in the gate run on the constraint update patch will actually run anything with those two together to make super sure it works
19:42:12 mriedem so you're thinking revert https://review.openstack.org/#/c/538541/ on stable/queens and depend on https://review.openstack.org/#/c/571540/ to get novaclient 10.1.0 so osc in grenade will use the osc proxy code that used to rely on novaclient,
19:42:34 mriedem i think the problem with that is the grenade revert won't run with the osc fix until that's released in https://review.openstack.org/#/c/571537/
19:42:59 mriedem i think you can hack around that,
19:43:11 mriedem by depending on a stable/queens devstack patch that installs osc from stable/queens source
19:43:19 melwitt yeah. I was thinking I could make that happen tho with a devstack DNM patch that does LIBS_FROM_GIT=python-openstackclient on branch stable/queens and then depends-on that
19:43:50 mriedem so you'd have something like, queens grenade revert -> depends-on -> devstack queens osc from source -> depends-on -> novaclient constraints update https://review.openstack.org/#/c/571540/
19:43:51 melwitt yeah exactly. just wanted to run it by you since your depends-on patch-verification-fu is strong
19:44:12 mriedem that's what i'd try i guess
19:44:36 mriedem OR
19:44:45 mriedem just get a queens devstack local and kick the tires
19:45:57 melwitt seeing something pass in the gate is more reassuring since I might screw up something in devstack somehow. but point taken
19:48:21 melwitt and also other people reviewing the changes can see it work, otherwise it's just "I tested it in devstack, you can trust me *wink*"
19:48:42 mriedem trust by verify?
19:48:47 mriedem i didn't know you were so into reagan
19:55:32 openstackgerrit Matt Riedemann proposed openstack/nova master: Remove support for /os-virtual-interfaces REST API https://review.openstack.org/569923
20:00:04 openstackgerrit Matt Riedemann proposed openstack/nova master: Remove support for /os-virtual-interfaces REST API https://review.openstack.org/569923
20:02:01 openstackgerrit Matt Riedemann proposed openstack/nova master: Remove support for /os-virtual-interfaces REST API https://review.openstack.org/569923
20:02:06 eandersson cfriesen, I see - yea was thinking of doing the same internally
20:02:33 mriedem eandersson: or, OR, you could work with the upstream community on digging ourself out of your hole
20:03:06 mriedem *yourself
20:04:35 eandersson haha yea that would probably be the more sensable thing to do =]
20:12:24 mriedem i'm trying hard to not make a starlingx comment here
20:13:26 mriedem if you called it rock-n-roll-racing-x then i could maybe buy into it
20:19:08 cfriesen mriedem: i'm probably corporately obligated to keep my mouth shut. :)
20:21:41 mriedem heh
20:21:50 mriedem it's not like huawei doesn't make changes
20:21:53 cfriesen mriedem: sorry, missed your earlier question. I was referring to providing CPU pinning (and hugepages) consumption data to an admin user via the API.
20:21:54 mriedem but i fight the same battle internally
20:22:18 mriedem cfriesen: sure but on the servers API or the hypervisors API?
20:22:50 cfriesen mriedem: ours is on the servers API, but we're private cloud and let people see more host-side info
20:23:15 mriedem well, you could do it on the servers API sure, we expose host/node etc, and that's controlled via policy
20:23:19 mriedem the same could be done here
20:23:41 mriedem someone just has to, you know, propose that api change in a spec first
20:24:17 mriedem artom: thanks for making those changes to the trusted vf series btw
20:24:45 cfriesen mriedem: http://paste.openstack.org/show/722471/ <-- an example of the stuff we added to "nova show"
20:26:11 artom mriedem, sure, thank you for finalizing your own nits :)
20:26:21 openstackgerrit Matt Riedemann proposed openstack/nova master: Match ComputeNode.uuid to ironic node uuid in RT https://review.openstack.org/571535
20:27:25 mriedem cfriesen_away: those don't seem crazy, and the nics thing is actually something that's been discussed a few times upstream, i.e. cleaning up GET /servers/{id}/ips and making the 'addresses' field in the existing response more useful
20:28:32 melwitt this is something where I feel like if we had enough info to put together a backlog spec, it might be something various contributors in the community might like to work on
20:30:05 mriedem reminds me of https://review.openstack.org/#/q/topic:521392+(status:open+OR+status:merged) which was approved, then needed a spec update, and seems to have been abandoned now
20:37:27 dansmith mriedem: I'm surprised that nics thing doesn't run contrary to your no-proxies leanings
20:39:07 openstackgerrit Matt Riedemann proposed openstack/nova master: Add osprofiler config options to generated reference https://review.openstack.org/571225
20:39:40 mriedem dansmith: i lean whichever way the winds of cash blow me
20:39:48 dansmith oh, well, then
20:40:50 mriedem we still have things like os-volume_attachments to show bdms attached to an instance, and os-interfaces to show ports attached to an instance,
20:41:03 mriedem and we've talked about leveraging those to return the tags associated with those devices
20:41:12 mriedem so we'll still have some things in the API as long as they are per-server
20:41:30 mriedem definitely not the stuff we had before, like CRUD ops on volumes/images/ports/networks
20:46:46 openstackgerrit Merged openstack/nova master: PowerVM Driver: vSCSI Fibre Channel volume adapter https://review.openstack.org/526094
20:46:53 openstackgerrit Merged openstack/nova master: PowerVM snapshot cleanup https://review.openstack.org/568988
20:47:08 dansmith volume attachments shows the mapping, but exposing things like the mtu is getting pretty specific
20:47:23 dansmith os-interfaces definitely exposes more stuff like that
20:48:12 dansmith but I dunno, I just don't know how to answer the question of why we don't have CRUD operations for all the values we expose from other services
20:48:52 dansmith I get the desire and need to be able to easily get all the IPs for the interfaces associated with the host, but not really network-specific things like mtu
20:48:59 dansmith s/host/instance/
20:52:50 melwitt dansmith: last patch for powervm merged so I've updated the runways, if you could do the honors updating the channel topic pls https://etherpad.openstack.org/p/nova-runways-rocky
20:54:27 melwitt thanks
20:55:17 dansmith yup
21:04:36 openstackgerrit Zack Cornelius proposed openstack/nova master: Implement file backed memory for instances in libvirt https://review.openstack.org/567876
21:09:16 cfriesen dansmith: personally I agree...I'd actually be fine with declaring that you have to pre-create ports in neutron and then pass them to nova. if desired the client could cross-reference with neutron for detailed info.
21:10:41 dansmith yeah, I also understand the network-id being there to assist in the mapping process, but anything that's not mapping should really be fetched from neutron I think
21:13:17 cfriesen in case it's not apparent, for dedicated cpus or multi-numa-node instances the "wrs-res:topology" line would have additional information
21:19:44 cfriesen what's the story with "meta.js" in the guest...is it still expected to work? I thought I remembered something about getting away from file injection.
21:28:37 openstackgerrit Matt Riedemann proposed openstack/nova master: Mention running rootwrap in daemon mode if hitting vif plug timeouts https://review.openstack.org/571577
21:34:03 openstackgerrit Julia Kreger proposed openstack/nova master: ironic: add instance_uuid before any other spawn activity https://review.openstack.org/563722
21:54:35 openstackgerrit Jay Pipes proposed openstack/nova master: placement: always create consumer records https://review.openstack.org/567678
21:54:36 openstackgerrit Jay Pipes proposed openstack/nova master: add consumers generation field https://review.openstack.org/557958
21:54:37 openstackgerrit Jay Pipes proposed openstack/nova master: placement: Allocation.consumer field https://review.openstack.org/565405
21:54:38 openstackgerrit Jay Pipes proposed openstack/nova master: rework allocation handler _allocations_dict() https://review.openstack.org/565407
21:54:39 openstackgerrit Jay Pipes proposed openstack/nova master: Add a microversion for consumer generation support https://review.openstack.org/565604
21:56:30 openstackgerrit Jan Gutter proposed openstack/os-vif master: Add multiqueue port profile for VIFGeneric https://review.openstack.org/570959
21:57:21 openstackgerrit Jan Gutter proposed openstack/nova master: Pass virtio multiqueue info to os-vif plugins https://review.openstack.org/571462
21:57:22 openstackgerrit Jan Gutter proposed openstack/nova master: Convert vrouter legacy plugging to os-vif https://review.openstack.org/571325
21:58:52 mriedem cfriesen: personality files (file injection) was deprecated from the api in queens
21:59:18 mriedem https://specs.openstack.org/openstack/nova-specs/specs/queens/implemented/deprecate-file-injection.html
21:59:56 mriedem cfriesen: and i assume you meant meta_data.json?
22:00:02 mriedem you just wouldn't get the 'files' key in there anymore
22:05:28 openstackgerrit Jay Pipes proposed openstack/nova master: Add a microversion for consumer generation support https://review.openstack.org/565604
23:11:00 openstackgerrit Merged openstack/nova stable/ocata: Avoid showing password in log https://review.openstack.org/561851
#openstack-nova - 2018-06-01
00:43:47 openstackgerrit wanghongtao proposed openstack/nova master: Fix the metadata re to match the unicode https://review.openstack.org/536236

Earlier   Later