| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-01-31 | |||
| 17:11:09 | bauzas | what's the context ? | |
| 17:11:32 | bauzas | btw. if anyone has some fakelibvirt expertise and functional testing for libvirt, raise a hand, please | |
| 17:11:48 | bauzas | it will prevent me to spend a couple of hours trying to understand the code | |
| 17:11:50 | stephenfin | bauzas: :) I'm looking for eyes on this backport https://review.openstack.org/#/c/530743/ | |
| 17:11:51 | mriedem | bauzas: https://review.openstack.org/#/c/538938/ | |
| 17:11:58 | mriedem | oops | |
| 17:12:00 | stephenfin | The stable/ocata backport too, but baby steps | |
| 17:12:04 | sean-k-mooney | gibi: yep that why at least one of the spec is written to assume nova will do the claim. i also have not really looked at them since november so cant rememeber all the deatils. | |
| 17:12:05 | openstackgerrit | sahid proposed openstack/nova-specs master: libvirt: add support for virtio-net rx/tx queue sizes https://review.openstack.org/539605 | |
| 17:12:24 | stephenfin | bauzas: I might be able to help | |
| 17:12:37 | stephenfin | w/ fakelibvirt, that is. What's up? | |
| 17:12:45 | bauzas | stephenfin: my goal is to provide a specific test module for vGPU testing | |
| 17:13:09 | bauzas | ie. mocking the virt driver to have mdevs and do a couple of server actions | |
| 17:13:23 | bauzas | like we have for NUMA or SR-IOV testing | |
| 17:13:46 | gibi | mriedem: done | |
| 17:13:50 | bauzas | stephenfin: but let's punt that discussion to tomorrow, /me looking at your change | |
| 17:13:56 | stephenfin | bauzas: Yup, that sounds fair. Have you looked at nova/tests/functional/libvirt/test_pci_sriov_servers.py ? | |
| 17:14:01 | bauzas | I did | |
| 17:14:09 | stephenfin | Ah, cool | |
| 17:14:40 | bauzas | I probably need to only identify the best entrypoints to mock only the pieces of libvirt I want | |
| 17:15:19 | stephenfin | Yeah, we can discuss tomorrow AM so, but fakelibvirt is basically the hardware-dependent aspects of libvirt mocked out. How those parts are determined, I'm not so sure | |
| 17:15:42 | stephenfin | vladikr would be the person to ask tomorrow, assuming he's about/has time | |
| 17:15:58 | bauzas | no worries, I'll dig into the code anyway | |
| 17:18:15 | dansmith | tssurya: assume you don't yet have a bug for the host_manager issue right? | |
| 17:18:20 | openstackgerrit | Marcin Juszkiewicz proposed openstack/nova master: Make sure that we have USB Host controller for USB tablet https://review.openstack.org/538003 | |
| 17:18:55 | hrw | mriedem, jaypipes: this version adds usb host controller ONLY if pointer device is 'usbtablet' so ps/2 mouse lovers are not touched | |
| 17:22:54 | mriedem | hrw: that needs tests | |
| 17:23:45 | tssurya | dansmith : no | |
| 17:23:49 | tssurya | do you want me to open one ? | |
| 17:24:19 | hrw | mriedem: I am aware. Will ask someone to write one for me as I know how get functionality working but completely out of ideas when it comes to tests. | |
| 17:24:20 | dansmith | tssurya: I was thinking maybe you already had bugs for both of these things when I asked on the first review, but then realized maybe not | |
| 17:24:38 | dansmith | tssurya: I think it'd be good to have the bugs coming from cern as backport justification, if you don't mind | |
| 17:25:01 | hrw | mriedem: having info are the idea and it's implementation ok is also helpful to me | |
| 17:25:06 | tssurya | dansmith : yea sure, I can do open them | |
| 17:25:15 | tssurya | you want it to be two seperate bugs ? | |
| 17:25:24 | mriedem | hrw: there should be at least one existing test in nova.tests.unit.virt.libvirt.test_driver that you could copy and modify | |
| 17:25:26 | dansmith | tssurya: yeah they're separate issues I think | |
| 17:25:40 | hrw | mriedem: ok | |
| 17:26:14 | hrw | mriedem: any other comments than unit testing? | |
| 17:26:31 | tssurya | cool will do and btw I have done the same for the bdm querying as we had discussed, for nova list, so will post that and open a bug for that too | |
| 17:26:54 | dansmith | tssurya: cool | |
| 17:27:01 | mriedem | hrw: haven't looked | |
| 17:27:05 | hrw | ok | |
| 17:27:14 | mriedem | don't wait for me though | |
| 17:27:23 | hrw | mriedem: do not plan to | |
| 17:27:34 | hrw | mriedem: and thanks a lot for reviews | |
| 17:27:40 | openstackgerrit | Elod Illes proposed openstack/nova master: Functional test: evacuate with no compute https://review.openstack.org/498482 | |
| 17:27:41 | bauzas | mriedem: tssurya: I missed your discussion about ReqSpec deletions | |
| 17:28:08 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add a nova-caching-scheduler job to the experimental queue https://review.openstack.org/539260 | |
| 17:28:14 | bauzas | mriedem: tssurya: I'll try to provide a new patchset this week, but no promises | |
| 17:29:00 | tssurya | bauzas : you have checked the last comment from mriedem on PS9 ? | |
| 17:29:07 | bauzas | not really | |
| 17:33:33 | bauzas | tssurya: dansmith: mmm, honestly, given https://review.openstack.org/#/c/515034/ I think we can abandon https://review.openstack.org/#/c/391060/ | |
| 17:33:35 | bauzas | mriedem: ^ | |
| 17:33:53 | bauzas | purging the spec records when we archive the deleted instances is enough | |
| 17:34:06 | dansmith | bauzas: no it's not | |
| 17:34:13 | bauzas | what are we missing then ? | |
| 17:34:14 | dansmith | for the case we were taking about above | |
| 17:34:34 | dansmith | because we wanted to use the deleted-reqspec as the sentinel to not include the instance in a list of instances from a downed cell | |
| 17:34:40 | dansmith | I want to talk about it in dublin | |
| 17:34:45 | dansmith | so don't abandon yet | |
| 17:34:48 | tssurya | bauzas : we are trying to find some way in which we can include information regarding the status of an instance : whether its deleted or not in the api DB | |
| 17:34:51 | bauzas | okay | |
| 17:37:39 | bauzas | dansmith: so the idea would be that if you don't find the related spec record for a specific instance UUID, then it implies that the instance is deleted, hence not needed to be shown from the instance mapping records ? | |
| 17:38:05 | bauzas | so we wouldn't have stale content if the cell is down? | |
| 17:38:06 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Check for leaked server resource allocations in post_test_hook https://review.openstack.org/538510 | |
| 17:38:14 | dansmith | bauzas: yeah | |
| 17:38:22 | bauzas | mmm, ok | |
| 17:38:26 | bauzas | I see | |
| 17:38:28 | dansmith | bauzas: we'd only be showing instance shells, but we don't want any to come back from the dead | |
| 17:38:39 | dansmith | we could have a deleted flag, but.. I really don't want to go down that road | |
| 17:38:48 | dansmith | if we can help it | |
| 17:38:58 | bauzas | the main problem is that I'm not sure we can deterministacally destroy a spec record | |
| 17:39:06 | bauzas | oh man | |
| 17:39:16 | bauzas | deterministically-something | |
| 17:42:33 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Check for leaked server resource allocations in post_test_hook https://review.openstack.org/538510 | |
| 17:43:17 | dansmith | bauzas: yeah I understand the potential issue there | |
| 17:43:18 | dansmith | hence the need to discuss | |
| 17:43:29 | bauzas | cool | |
| 17:43:50 | bauzas | okay, let's keep that open until Dublin then | |
| 17:43:52 | openstackgerrit | Dan Smith proposed openstack/nova master: Make host_manager use scatter-gather and ignore down cells https://review.openstack.org/539617 | |
| 17:43:58 | bauzas | dansmith: thanks for explaining me the need | |
| 17:44:01 | dansmith | tssurya: ^ I'll update with the bug number when you get it | |
| 17:46:57 | hrw | hm. I am reading _get_guest_pointer_model() again and again... It returns anything ONLY if that anything is usbtablet... | |
| 17:47:17 | hrw | or I do not understand Python (which is highly possible) | |
| 17:47:37 | bauzas | mriedem: before leaving, just left +1 on https://review.openstack.org/#/c/530743/ because I need to consider whether it's acceptable to have exceptions or not that are popped by the API, based on the compute version | |
| 17:48:25 | bauzas | like, say I'm upgrading half of my nodes, is that a problem if I'm getting a different disk calculation logic ? | |
| 17:48:33 | bauzas | either way, the master change is merged | |
| 17:48:47 | bauzas | so maybe it's just me overthinking | |
| 17:49:04 | bauzas | anyway, need to disappear | |
| 17:50:18 | hrw | function starts. then hw_pointer_model is read from image properties. if it is not set then check for use_usb_tablet option and if it is set then set pointer_model to usbtablet and issue a warning that configuration needs update. Then is a check is pointer_model a usbtablet. If it is not then function ends. It it is then check for is Vnc or (is spice but without spice agent) and then return usbtablet | |
| 17:50:24 | hrw | through _get_guest_usb_tablet() function which checks for VMMode and creates structure for add_device() call | |
| 17:52:26 | hrw | mriedem: added one more comment and ending a day | |
| 17:53:12 | tssurya | dansmith : https://bugs.launchpad.net/nova/+bug/1746558 , https://bugs.launchpad.net/nova/+bug/1746561 | |
| 17:53:14 | openstack | Launchpad bug 1746558 in OpenStack Compute (nova) "Make service all-cells min version helper use scatter-gather" [Undecided,New] - Assigned to Dan Smith (danms) | |
| 17:53:15 | openstack | Launchpad bug 1746561 in OpenStack Compute (nova) "Make host_manager use scatter-gather and ignore down cells" [Undecided,New] - Assigned to Dan Smith (danms) | |
| 17:53:32 | tssurya | dansmith: thanks again! | |
| 17:53:58 | openstackgerrit | Dan Smith proposed openstack/nova master: Make service all-cells min version helper use scatter-gather https://review.openstack.org/539590 | |
| 17:53:59 | openstackgerrit | Dan Smith proposed openstack/nova master: Make host_manager use scatter-gather and ignore down cells https://review.openstack.org/539617 | |
| 17:54:01 | dansmith | tssurya: they haven't started running tests yet, so ^ | |