Earlier  
Posted Nick Remark
#openstack-nova - 2018-05-10
11:05:37 openstackgerrit Yikun Jiang (Kero) proposed openstack/nova master: WIP: Microversion 2.63 - Use new format policy in server group https://review.openstack.org/567534
11:11:11 openstackgerrit Chris Dent proposed openstack/nova master: VMware: Live migration of instances https://review.openstack.org/270116
11:55:23 bhagyashri_s cdent: Hi, I am trying to write the functional test case for change https://review.openstack.org/#/c/560459 . I have written functinal test for the change # link http://paste.openstack.org/show/720747/ in /nova/tests/functional/test_servers.py
11:56:58 cdent hi bhagyashri_s. is it not doing what you want?
12:02:04 cdent bhagyashri_s: for what you are trying to test, a functional test in test_servers is probably not what you want
12:03:09 cdent because what you're trying to test is that the virt driver is creating the right data (in the ProviderTree), not create resource providers
12:03:26 cdent bhagyashri_s: basically what Eric says in his comment
12:10:18 bhagyashri_s cdent: Firstly the change is regarding the shared resource provider thing . When user configure the shared resource provider then at that time the DISK_GB inventory should report from shared rp and not from the compute node rp
12:11:41 bhagyashri_s cdent: and for that libvirt driver update_provider_tree() removes the DISK_GB inventory from the cn rp if the shared rp is configure
12:12:38 cdent so in your test you should need to create a server. instead you just need to confirm that the correct resource providers and inventory are created
12:12:49 cdent sorry: shouldn't need!
12:14:47 bhagyashri_s cdent so in my functional test case i am trying to update by calling the resource tackers update_provider_tree which basically updates the inventory internally using def _update() method which calls the drivers update_provider_tree() method and in my test case how should i instantiate the LibvirtDriver
12:17:06 openstackgerrit Petersingh Anburaj proposed openstack/nova master: Making consistent used of GiB and MiB in Doc https://review.openstack.org/559985
12:18:53 cdent bhagyashri_s It sounds like you are trying to test from too high a level. Maybe try: Set up the sharing resource provider and put it an aggregate. Then create a ProviderTree representing a compute node, put it in aggregate too, then pass that to the method on the virt driver and make sure that the provider tree that comes back doesn't have local disk. I'm not sure of the setup you'll need to do for that, without doing a
12:18:54 cdent more digging
12:19:19 cdent if/when efried comes back you might check with him, but the virt drivers are an area I've done very little work with
12:23:13 openstackgerrit Nguyen Hai proposed openstack/nova-specs master: Follow the new PTI for document build https://review.openstack.org/551802
12:23:21 jroll someone wake up mriedem before I forget to tell him a thing :P
12:25:01 cdent destroy all irc
12:27:19 efried ō/
12:27:20 bhagyashri_s cdent: ok ty
12:27:37 efried reading back...
12:33:44 efried cdent, bhagyashri_s: Yes we want a unit test, but we also want a functional test. The functional test should do just what your paste shows. When we talked about this a couple of days ago, I offered to help you debug it IF you posted your WIP test case to the change set so I could download it and be at the same baseline, rather than copy/pasting from pastebin and trying to figure out where to put it and how to run it.
12:37:54 bhagyashri_s efried: because of some environmental issue iam not able to push patches so just paste the code in paste.openstack.org
12:42:13 bhagyashri_s efried: once it will resolved i will push it, I catch the issue in my functional test case, as the change is in libvirt code so through the functional test case the driver is not loaded and that's why the inventory is not properly updated so my doubt is how i will initiate the driver so that it report the correct inventory if the shared resource provider is configured
12:43:40 bhagyashri_s efried: sorry for inconvenience i am not able to push the patch
12:43:41 bauzas FWIW, folks I'm on PTO till end of the week
12:43:52 bauzas call it May-hem
12:44:19 efried bhagyashri_s: okay, can you pastebin the whole file including imports and tell me the path where you put it?
12:47:06 bauzas efried: jaypipes: cdent: I'm on slides mode for my YVR talk but something came out of my mind. Do we plan to have quota resources using Placement resource classes ?
12:47:24 bauzas efried: jaypipes: cdent: for example, we now have VGPU RC
12:47:39 bauzas efried: jaypipes: cdent: I guess operators would love to get quotas for vGPUs
12:47:50 bauzas so, how to help them ?
12:47:55 efried I still don't know what a quota is
12:48:13 bauzas efried: it's old story
12:48:30 bauzas efried: https://docs.openstack.org/nova/pike/admin/quotas.html
12:48:53 bhagyashri_s efried: #link http://paste.openstack.org/show/720752/ path: nova/tests/functional/test_servers.py
12:50:50 bauzas efried: and https://docs.openstack.org/nova/latest/user/quotas.html
12:53:24 openstackgerrit Hironori Shiina proposed openstack/nova master: ironic: Support resize and cold migration https://review.openstack.org/500677
13:05:26 openstackgerrit Lee Yarwood proposed openstack/nova master: Rename parameters in call() of nova/image/glance.py https://review.openstack.org/508533
13:17:17 moshele melwitt: hi, the cinder changes for nvmeof is merged in cinder so can you review the nova change https://review.openstack.org/#/c/482640/
13:21:31 efried bauzas: So you're thinking about implementing quotas in placement, such that a user and/or project ID is allowed only a certain amount of a given resource class?
13:22:03 bauzas efried: well, yes
13:22:44 efried bauzas: It seems like a reasonable idea. I don't recall having heard of any plans along those lines, but that doesn't mean it's not on the radar for someone.
13:22:57 efried bauzas: queue it up for Denver?
13:23:13 bauzas efried: my point is about the fact we created a new resource class, VGPU
13:23:21 bauzas efried: now the scheduler verifies it
13:23:24 jaypipes bauzas: that's a good discussion. I'd like to have melwitt in it, though.
13:23:29 efried agree
13:23:30 jaypipes bauzas: shall we wait for her?
13:23:42 bauzas jaypipes: sure, but I'm in vacations atm :)
13:23:46 bauzas in the beach :)
13:24:07 bauzas so when she's here, not sure I'll be here too :p
13:24:12 bauzas but anyway
13:24:19 bauzas nothing really important atm
13:24:24 bauzas it's more for my talk
13:24:24 jaypipes bauzas: alex_xu was also working on quotas and placement things
13:24:40 bauzas gtk
13:24:53 bauzas AFAIK, we don't support custom quotas, right?
13:26:08 jaypipes bauzas: I'm not sure any more.
13:26:14 bauzas yup
13:26:27 bauzas it was possible previously
13:26:40 bauzas but when we provided the quotas for cells v2, then we stopped that
13:26:54 bauzas that's what I remember at least
13:31:10 jaypipes bauzas: right. I was under the impression that we (nova) would start using the placement API for various usage queries. but I don't think that's been done yet.
13:35:56 jaypipes Kevin_Zheng: lmao 'All of a sudden, everyone uses the word "granular"'
13:36:02 jaypipes efried: ^
13:38:15 mriedem stephenfin: bauzas: can one of you take a look at this? https://review.openstack.org/#/c/566367/ needs a final +2 and then we need to backport it through to ocata
13:38:39 bauzas mriedem: I'm on PTO today and tomorrow, but see if I can help
13:43:21 jroll jaypipes: bauzas: yeah, I know it's on alex_xu's list to do quotas on resource classes, not sure if he has anything proposed upstream yet
13:43:45 stephenfin mriedem: Sure
13:43:56 openstackgerrit Stephen Finucane proposed openstack/nova master: objects: Add NUMANetworkInfo object https://review.openstack.org/564439
13:43:57 openstackgerrit Stephen Finucane proposed openstack/nova master: conf: Add '[neutron] physnets' and related options https://review.openstack.org/564440
13:43:58 openstackgerrit Stephen Finucane proposed openstack/nova master: libvirt: Start populating NUMACell.network_info field https://review.openstack.org/564441
13:43:59 openstackgerrit Stephen Finucane proposed openstack/nova master: objects: Add NetworkRequest.physnet, tunnelled https://review.openstack.org/564442
13:44:00 openstackgerrit Stephen Finucane proposed openstack/nova master: network: Unchain _get_phynet_info from _get_port_vnic_info https://review.openstack.org/564443
13:44:01 openstackgerrit Stephen Finucane proposed openstack/nova master: network: Add 'populate_networks' to network API https://review.openstack.org/564444
13:44:02 openstackgerrit Stephen Finucane proposed openstack/nova master: network: Add '_get_network_tunnel_status' https://review.openstack.org/564445
13:44:03 openstackgerrit Stephen Finucane proposed openstack/nova master: db: Add network_requests to instance_extra https://review.openstack.org/564446
13:44:04 openstackgerrit Stephen Finucane proposed openstack/nova master: objects: Add Instance.network_requests https://review.openstack.org/564447
13:44:05 openstackgerrit Stephen Finucane proposed openstack/nova master: hardware: Start accounting for network requests https://review.openstack.org/564448
13:44:06 openstackgerrit Stephen Finucane proposed openstack/nova master: rt: Start utilizing Instance.network_requests https://review.openstack.org/564449
13:44:07 openstackgerrit Stephen Finucane proposed openstack/nova master: objects: Add RequestSpec.requested_networks https://review.openstack.org/564450
13:44:08 openstackgerrit Stephen Finucane proposed openstack/nova master: api: Remove unnecessary default parameter https://review.openstack.org/564451
13:44:09 openstackgerrit Stephen Finucane proposed openstack/nova master: scheduler: Start utilizing network_requests https://review.openstack.org/564452
13:44:12 jaypipes jroll: last I heard, he was focusing on quota-by-aggregate and ran into a number of issues...
13:44:29 jroll jaypipes: yar, he's working through it
13:44:31 jaypipes jroll: sorry, had to wait until stephenfin's patch-cloud passed.
13:44:35 jroll heh
13:44:47 stephenfin ☁ ☁ ☁
13:44:56 jaypipes which reminds me I need to complete reviews on those. :)
13:45:36 stephenfin jaypipes: Would be appreciated. Fixed the Tempest issue (thanks gibi/mriedem) and addressed what comments were there so far
13:45:41 jaypipes gibi: I've read your ML post about the instance migration UUID thing three times and looked at the map_instances() function. and I still don't get it :(
13:46:43 stephenfin jaypipes: we're using a munged UUID as a marker so we catch batch up the migrations
13:47:03 stephenfin cos it's munged, it's not a valid UUID and oslo.whatever complains
13:47:35 jaypipes stephenfin: right, but I don't understand why we're doing the replace('-','') thing at all
13:47:36 stephenfin edleafe tried to fix it before but dansmith found the solution distressing
13:47:51 stephenfin because there's a UNIQUE constraint on that column
13:48:07 stephenfin if we didn't do that, it would clash with the instance (or whatever that UUID is taken from)
13:48:08 jaypipes stephenfin: ok, but why are we changing it?

Earlier   Later