| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2019-01-30 | |||
| 16:39:51 | openstackgerrit | Matthew Booth proposed openstack/nova master: Fix SSL infinite recursion running WSGI app under python 3.7 https://review.openstack.org/626952 | |
| 16:40:32 | lyarwood | mdbooth: tbh it was a random punt at the underlying problem so I just used related, it should be close now tbh | |
| 16:41:05 | mdbooth | lyarwood: Yeah, that's what I thought too. I updated. | |
| 17:58:29 | mriedem | stack@stein:~/nova$ curl http://162.253.55.42/compute/v2.1 | |
| 17:58:29 | mriedem | oh fun | |
| 17:58:30 | mriedem | Out of date osapi_compute service This service is older (v36) than the minimum (v37) version of the rest of the deployment. Unable to continue. | |
| 17:58:42 | mriedem | with all services in nova_cell0 and nova_cell1 being version 37 | |
| 17:59:01 | mriedem | which sharts out 500s from the api when listing services | |
| 18:00:08 | mriedem | hell i'm not even listing services there, just getting the version doc | |
| 18:00:09 | mriedem | wtf | |
| 18:00:37 | mgagne | what's the purpose of reservations in Nova? can they all be deleted without impact? I have a situation where *some* reservations where not expired as expected and I'm now stuck with them. | |
| 18:00:47 | mriedem | mgagne: depends on the release | |
| 18:00:58 | mriedem | reservations aren't used starting in pike | |
| 18:01:02 | mgagne | mriedem: mitaka | |
| 18:01:04 | mriedem | they are related to quota | |
| 18:01:35 | mgagne | mriedem: with cellsv1 of course. they are currently located in the API cell database. | |
| 18:02:17 | mriedem | they are meant to be short-term | |
| 18:02:29 | mgagne | right. | |
| 18:02:34 | mriedem | they are a placeholder for quota until a quota 'claim' is either committed or rolled back | |
| 18:02:57 | mriedem | so if the server a given reservation is already gone or active, you don't need the reservation anymore | |
| 18:03:01 | mgagne | I'm wondering what deletes/expires them. I read somewhere it's the scheduler but I'm wondering in which case they get expired/deleted so I can see if I can flush them or not. | |
| 18:03:25 | mriedem | there is also a periodic task that reaps them i think, check for the "reservation_expire" option | |
| 18:03:53 | mriedem | anyway, i *think* (without looking at code) reservations are meant to be deleted once the quota claim is committed or rolled back | |
| 18:03:58 | mgagne | it's a custom patch for cellsv1 which I cherry-picked. But dumb me didn't set it to true. | |
| 18:04:08 | mriedem | reservation_expire is a timer | |
| 18:04:34 | mgagne | hmm, ok, I think it's cells/expire_reversations I had | |
| 18:05:13 | mgagne | ok, will look into the claim thing and see what I can do. there is just too many rows to filter them, it's gonna take forever to complete. | |
| 18:05:22 | mriedem | i don't see that in https://docs.openstack.org/nova/rocky/configuration/config.html | |
| 18:05:57 | mgagne | https://github.com/NeCTAR-RC/nova/commit/6a5a4e53a008ab4032917a9e682e32df1279d758 | |
| 18:06:43 | mriedem | https://github.com/openstack/nova/blob/newton-eol/nova/cmd/manage.py#L306 existed in newton | |
| 18:06:52 | mriedem | you could check that, i'm not sure if it cleans up old reservations though | |
| 18:07:36 | mriedem | ah ok, well that's out of tree | |
| 18:07:44 | mriedem | sorrison is your man | |
| 18:08:25 | mgagne | yea, I just need to find a way to clear up my hmmm amount of reservations. | |
| 18:16:18 | mriedem | well this isn't good, "Updating service version for nova-osapi_compute on stein from 37 to 36" | |
| 18:16:21 | mriedem | on restart of n-api | |
| 18:17:00 | cdent | ruh roh | |
| 18:58:50 | mriedem | think i figured it out | |
| 19:03:50 | mriedem | i stacked with current master, then cherry picked an older change that needed to be rebased which backleveled the service version from 37 to 36 and then when restarting n-api/n-cpu they blew up (well, n-cpu blew up, n-api just logged it) | |
| 19:13:50 | openstackgerrit | Kashyap Chamarthy proposed openstack/nova-specs master: WIP: Gracefully handle QEMU machine types for guests https://review.openstack.org/631154 | |
| 19:38:18 | mriedem | efried: alex_xu: regarding the shared snapshot image patch https://review.openstack.org/#/c/630769/ i tested that with devstack again and alex_xu is correct that when the non-admin project unshelves (or deletes) the server they don't have access to delete the image snapshot owned by the admin project, but it doesn't make either operation fail, it just logs a warning | |
| 19:38:22 | mriedem | i think that is OK | |
| 20:58:00 | cfriesen | stephenfin: sean-k-mooney: does https://github.com/openstack/nova-specs/blob/master/specs/queens/implemented/share-pci-between-numa-nodes.rst apply to SR-IOV and PCI-PT NICs? | |
| 20:59:46 | efried | mriedem: ack | |
| 21:00:07 | efried | Is a(ny) failure during spawn reschedulable? | |
| 21:00:24 | efried | (unrelated ^ ) | |
| 21:07:27 | mriedem | efried: no | |
| 21:07:41 | mriedem | for example, anything out of _prep_block_devices that fails is an abort | |
| 21:07:50 | mriedem | oh spawn.. | |
| 21:08:39 | mriedem | depends on what it is https://github.com/openstack/nova/blob/master/nova/compute/manager.py#L2120 | |
| 21:08:54 | mriedem | e.g. these don't get rescheduled https://github.com/openstack/nova/blob/master/nova/compute/manager.py#L2173 | |
| 21:09:33 | mriedem | several more in the block below | |
| 21:09:38 | mriedem | anything not specifically caught is rescheduled | |
| 21:09:56 | efried | mriedem: Got it, thanks. | |
| 21:12:23 | sean-k-mooney | cfriesen yes and it used to have have both flavor extra specs and an image property to allwo you to set the policy | |
| 21:12:29 | sean-k-mooney | cfriesen: https://github.com/openstack/nova-specs/blob/f350472e627db1afbe5e32c35a9973124642ecf1/specs/queens/implemented/share-pci-between-numa-nodes.rst#proposed-change | |
| 21:12:41 | sean-k-mooney | cfriesen: but it was not implemanted and the spec was retroactivly updated. | |
| 21:13:11 | sean-k-mooney | cfriesen: i want to add those back in in train and complete the original spec | |
| 21:13:31 | sean-k-mooney | cfriesen: i discoverd this in novemebr but did not have time to work on it this cycle | |
| 21:13:41 | cfriesen | sean-k-mooney: right, I remember that. so how does the pci alias in the config file map to what gets used for SR-IOV? | |
| 21:14:07 | sean-k-mooney | cfriesen: the alias works assuming stepens patch merged | |
| 21:14:22 | sean-k-mooney | cfriesen: there was a bug were we were not storing the policy in the db | |
| 21:14:45 | cfriesen | sean-k-mooney: as I recall, we decided that setting the policy on the alias was sufficient since most likely all requests for a particular device would use the same PCI affinity (ie either you care about performacne on that device or not) | |
| 21:15:21 | sean-k-mooney | cfriesen: we need these https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:master+topic:bug/1805891 | |
| 21:15:42 | sean-k-mooney | cfriesen: https://bugs.launchpad.net/nova/+bug/1805891 | |
| 21:15:44 | openstack | Launchpad bug 1805891 in OpenStack Compute (nova) "pci numa polices are not followed" [High,In progress] - Assigned to Stephen Finucane (stephenfinucane) | |
| 21:15:47 | cfriesen | sean-k-mooney: we don't specify an alias explicitly for SR-IOV networking though, right? it looks up the network in the pci whitelist. So how do we decide which alias to use? | |
| 21:16:42 | sean-k-mooney | oh right sorry ya there is no way to adjust the policy for srivo devices at all | |
| 21:17:00 | sean-k-mooney | i taught the policy was in teh whitelist for a second not the alais | |
| 21:17:18 | cfriesen | on reflection, that kind of seems like an oversight. :) | |
| 21:17:33 | sean-k-mooney | cfriesen: well it was in the spec for a reason | |
| 21:17:59 | sean-k-mooney | specifcally so that neutron sriov prot could be supported | |
| 21:18:32 | cfriesen | I think when we switched to putting it in the alias we probably missed the fact that the alias didn't apply to devices associated with neutron ports | |
| 21:19:14 | cfriesen | certainly I didn't realize it at the time | |
| 21:19:43 | sean-k-mooney | well the alias was never ment to be suported at all | |
| 21:20:55 | sean-k-mooney | i personally think we shoudl have followed the spec in the frist place. | |
| 21:24:20 | sean-k-mooney | cfriesen: anyway am im hoping to change this in train. i think i commented on the starlingx tread a few montsh ago about the histroy of this and the current state | |
| 21:24:56 | cfriesen | sean-k-mooney: okay, will keep an eye out | |
| 22:09:37 | artom | https://review.openstack.org/#/c/599208/ | |
| 22:09:56 | artom | Please someone tell me I'm not the only struggling with almost every single line in that beast | |
| 22:14:24 | efried | artom: I admit I haven't looked at it again since a much earlier PS. It's on my list, and steadily creeping higher. | |
| 22:16:09 | mriedem | artom: i just realized that these bdm docs don't mention tags at all https://docs.openstack.org/nova/latest/user/block-device-mapping.html#block-device-mapping-v2 | |
| 22:21:10 | openstackgerrit | Merged openstack/nova master: Consolidate inventory refresh https://review.openstack.org/615695 | |
| 22:21:17 | efried | woot | |
| 22:28:39 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Add microversion to expose virtual device tags https://review.openstack.org/631948 | |
| 22:53:17 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Fix config docs for handle_virt_lifecycle_events https://review.openstack.org/634058 | |
| 23:09:45 | artom | mriedem, yeah, except the api-ref there's 0 device tagging documentation | |
| 23:10:02 | artom | It's been a low-ish priority item on my backlog for a long time | |
| 23:10:34 | artom | I wrote https://notartom.net/2017/06/20/virtual-device-role-tagging-better-explained/ but need to "upstream" it properly | |
| 23:56:03 | openstackgerrit | Takashi NATSUME proposed openstack/nova master: Remove deprecated 'os-flavor-manage' policy https://review.openstack.org/633656 | |
| 23:56:17 | openstackgerrit | Takashi NATSUME proposed openstack/nova master: Fix a missing policy in test policy data https://review.openstack.org/633686 | |
| #openstack-nova - 2019-01-31 | |||
| 00:02:33 | openstackgerrit | Takashi NATSUME proposed openstack/nova master: Remove mox in unit/network/test_neutronv2.py (3) https://review.openstack.org/574104 | |
| 00:02:45 | openstackgerrit | Takashi NATSUME proposed openstack/nova master: Remove mox in unit/network/test_neutronv2.py (4) https://review.openstack.org/574106 | |
| 00:02:57 | openstackgerrit | Takashi NATSUME proposed openstack/nova master: Remove mox in unit/network/test_neutronv2.py (5) https://review.openstack.org/574110 | |
| 00:03:07 | openstackgerrit | Takashi NATSUME proposed openstack/nova master: Remove mox in unit/network/test_neutronv2.py (6) https://review.openstack.org/574113 | |
| 00:03:19 | openstackgerrit | Takashi NATSUME proposed openstack/nova master: Remove mox in unit/network/test_neutronv2.py (7) https://review.openstack.org/574974 | |
| 00:03:30 | openstackgerrit | Takashi NATSUME proposed openstack/nova master: Remove mox in unit/network/test_neutronv2.py (8) https://review.openstack.org/575311 | |
| 00:08:26 | alex_xu | mriedem: thanks. so we just need to ask admin to cleanup the image after user unshelve the instance | |
| 00:08:41 | alex_xu | that seems fine | |
| 00:16:37 | openstackgerrit | Hang Yang proposed openstack/nova stable/queens: Fix port dns_name reset https://review.openstack.org/633807 | |