Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-02-28
16:11:24 lyarwood akijak: hey, have you looked at https://github.com/openstack/nova/blob/master/nova/compute/manager.py#L2679 and how we are passing image_meta around?
16:11:31 mriedem efried: i can create one
16:11:48 efried mriedem Thanks. And wipe the agenda for Thursday's meeting?
16:11:59 mriedem efried: it's only tuesday!
16:12:09 mriedem i get to the thurs meeting agenda by about wed night
16:12:19 efried Heh. Didn't know how much lead time we were supposed to have ;-)
16:12:23 mriedem efried: if you need to change something in the meeting agenda, you can do that now
16:12:29 efried rgr
16:14:27 spotz mriedem: 0 instances in Error, same result
16:15:15 dansmith mriedem: so I noticed some weird quota behavior when I was trying to get my multicell patch working
16:15:42 dansmith spotz: what is the [database]/connection set to in the nova conf that you're using for your apis?
16:15:49 dansmith and conductors
16:16:01 dansmith maybe we're reserving in cell0
16:17:00 mriedem i assume it's mysql+pymysql://nova:linuxacademy123@controller/nova
16:17:04 mriedem from https://gist.github.com/spotz/42b9d15f016073458742268735644578
16:18:16 dansmith actually, it has to be or other things won't work because ocata
16:18:16 spotz mriedem: Yep
16:19:04 spotz I know the basic config docs that are there minus placement-api and cells work as I also built a newton cluster and it's happy
16:19:24 spotz We're missing something, whether it's a step or something out of order
16:23:09 efried mriedem: I can start populating https://etherpad.openstack.org/p/pike-nova-priorities-tracking with the structure of its ocata predecessor if you like.
16:23:29 efried You have more important (or at least more urgent) things to do ;-)
16:24:03 mriedem efried: yes please
16:24:07 mriedem i'm doing 3 things at once
16:26:47 mriedem spotz: can you make a curl request to the placement endpoint using the same creds as are in the nova.conf on the compute?
16:27:04 spotz mriedem: Yep hang on
16:27:24 mriedem to /resource_providers would be enough
16:36:38 spotz mriedem: Bad Request but it could have been my curl
16:49:40 spotz mriedem: Yeah everytime I try to curl I'm getting the 400, I'm assuming you wanted me to just get like tokens?
16:50:48 mriedem spotz: just get the /resource_providers
16:50:51 mriedem you'll need a token
16:51:48 mdbooth mriedem: https://review.openstack.org/#/c/395709/
16:52:05 mriedem mdbooth: you don't need that
16:52:06 mdbooth Nothing unexpected in the logs, either
16:52:16 mriedem we run the nfs job in the nova experimental queue
16:52:16 mdbooth mriedem: You mean the check experimental job?
16:52:27 mdbooth Yeah, we run a bunch of other stuff in there too, though
16:52:30 mriedem "check experimental" runs the set of experimental jobs on your nova change
16:52:31 mriedem directly
16:52:44 mdbooth That one only runs the nfs job, so it's kinda nice
16:54:59 mdbooth mriedem: check experimental is a bit of a nose bleed. How much of it is intended to be red vs green?
16:55:14 mriedem i need to remove some of it
16:55:25 mriedem ala https://review.openstack.org/#/c/415033/
16:55:31 mriedem but time is an issue
16:56:15 mdbooth mriedem: The nfs job it ran passed
16:57:03 mdbooth mriedem: Incidentally, would it be practical to do 'check
16:57:03 mdbooth gate-tempest-dsvm-full-devstack-plugin-nfs-nv', and have it run just that job?
16:57:28 mriedem spotz: http://paste.openstack.org/show/600806/
16:57:36 mriedem ^ is from my devstack, with curl to placement
16:57:50 mriedem mdbooth: that's not a thing
16:57:51 mdbooth Or have a much wider array of 'interest areas' maybe, like 'check nfs', or 'check storage'
16:58:08 mdbooth mriedem: Right, but is there any reason it's not a thing?
16:58:33 mriedem someone would have to do it
16:58:47 mdbooth Sure. Would it be a useful thing?
16:58:52 mriedem idk
16:59:06 mdbooth For eg, in this case we've run gate-tempest-dsvm-lxc-ubuntu-xenial, and it failed, but I don't really care.
16:59:31 mriedem what would have been useful was if the hilton didn't f up charging me for last week so my internal expense reporting tool took over an hour to trick to make up for a separate .05 charge
16:59:51 mriedem recheck hilton
16:59:55 mdbooth lol
17:00:28 mriedem spotz: i get the token via: OS_TOKEN=$(openstack token issue -c id -f value)
17:00:36 mdbooth I'm pretty sure that the scope of expense reporting tools vastly exceeds anything we have to deal with :)
17:01:57 spotz mriedem: Still getting no json object
17:02:12 mriedem spotz: but a 200 or an error?
17:03:25 spotz mriedem: 200, I think it's me I suck at curl
17:03:54 mriedem and the creds you're using for that token are the same as what's in nova.conf on the compute?
17:04:44 spotz mriedem: Yeah I set up an rc and did an openstack token issue
17:05:04 spotz Or I could be doing things ass backwards
17:05:22 mriedem are you still seeing warnings in the nova-compute logs about not being able to connect to placement?
17:06:42 spotz mriedem: Nope
17:09:58 mriedem but there are no resource providers coming back...
17:10:03 mriedem something is janked there then
17:10:17 mriedem because the compute is running a periodic task every minute that will update it's info in the placement api
17:10:30 mriedem if it can't, you should be seeing a warning every 10 minutes or so in the compute logs
17:11:17 spotz nova-compute.log is now lines of 017-02-28 11:05:39.099 2618 INFO nova.compute.resource_tracker [req-8025f2c1-a990-4383-ab04-46e929ffce39 - - - - -] Final resource view: name=compute phys_ram=7983MB used_ram=512MB phys_disk=5GB used_disk=0GB total_vcpus=1 used_vcpus=0 pci_stats=[]
17:11:33 mriedem that's not it
17:11:34 cfriesen with the change to make live migration checks async in Newton, did horizon get updated to properly report any errors in the checks?
17:11:57 mriedem cfriesen: it's a microversion so horizon would have to be opting in
17:11:57 mriedem and if so,
17:11:58 mriedem they have to adjust for the behavior change
17:12:43 openstackgerrit John Griffith proposed openstack/nova master: Add attachment_uuid to BDM table https://review.openstack.org/438744
17:13:59 openstackgerrit John Griffith proposed openstack/nova master: Add Cinder V3 Detach calls https://review.openstack.org/438750
17:14:01 openstackgerrit STEW TY proposed openstack/nova master: Complete verification of servers-action-fixed-ip.inc https://review.openstack.org/338606
17:14:16 mriedem sfinucan: mdbooth: you were asking why lyarwood was adding an attachment_id to the bdm table
17:14:17 mriedem https://review.openstack.org/#/c/373203/
17:14:18 mriedem is why
17:14:24 mriedem you should read it since we need more eyes on it
17:15:35 mriedem dansmith: what do you think, should we shit or get off this pot? https://review.openstack.org/#/c/242602/
17:15:47 dansmith mriedem: lol
17:16:01 mriedem we have no immediate need for it right now that i'm aware of
17:16:05 mdbooth mriedem: Coincidentally just reading that
17:16:15 mriedem but it's always something that comes up as "if we only had a uuid on the bdm"
17:16:16 dansmith mriedem: I don't care.. seems like a good thing, and people seem to occasionally need it
17:16:17 dansmith but yeah if no need then whatever
17:16:20 mdbooth lyarwood: You still need a commit message btw :P
17:16:23 dansmith I can abandon if it's bothering you
17:16:29 mriedem it's not bothering me
17:16:35 mriedem but lyarwood is building on it
17:16:35 lyarwood mdbooth: meh, can't you read my mind?
17:17:06 mdbooth lyarwood: Not until you push it to gerrit :)
17:17:11 mriedem lyarwood: consequently, what state is https://review.openstack.org/#/c/290793/ in now?
17:19:08 lyarwood mriedem: rotting on the floor, I had an open question around how to actually use the new UUIDs to avoid this iirc

Earlier   Later