| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-03-31 | |||
| 17:55:36 | sdague | it has a lot of other implications, he's going to write the whole thing up to share on the ML | |
| 17:55:46 | mriedem | ok | |
| 17:55:50 | sdague | but, I wanted to get that seed out there for everyone to ponder | |
| 17:56:18 | sdague | it would mean we'd start testing some of the pathways that aren't otherwise tested | |
| 17:56:35 | mriedem | well, some of those are things like numa/pci right? | |
| 17:56:45 | mriedem | which i don't think we'd be testing regardless | |
| 17:57:24 | clarkb | I think there are newer live migration things too | |
| 17:57:36 | clarkb | at least my tests tripped over that with bad wheel builds against old libvirt | |
| 17:58:03 | sdague | yeh, postcopy | |
| 17:58:23 | sdague | and virtlogd | |
| 17:58:39 | mriedem | what are we right now? 1.3.1? | |
| 17:58:43 | sdague | yep | |
| 17:59:15 | sdague | and perf_events? | |
| 17:59:33 | mriedem | yeah, so the only thing of those that would get tested by tempest are probably postcopy | |
| 17:59:42 | mriedem | unless virtlogd ties in with any console-related tests | |
| 17:59:45 | sdague | that seems like everything with higher versions bring in | |
| 18:00:35 | mriedem | yeah the perf events stuff won't be tested by tempest directly | |
| 18:00:36 | sdague | yeh, I think there might be console tests there | |
| 18:00:52 | sdague | anyway, a thing to ponder, details comming. | |
| 18:02:19 | mriedem | https://www.youtube.com/watch?v=09kuOAr_wkc | |
| 18:02:24 | mriedem | i'll ponder it on the tree of woe | |
| 18:02:58 | openstackgerrit | David Moreau Simard proposed openstack/nova master: Be honest with users when no valid host was found https://review.openstack.org/452279 | |
| 18:03:12 | sdague | heh | |
| 18:11:04 | leakypipes | sdague, sfinucan: another easy one: https://review.openstack.org/#/c/447401/5 | |
| 18:11:39 | sdague | bam | |
| 18:11:53 | leakypipes | heh | |
| 18:20:23 | mriedem | i need another specs core on https://review.openstack.org/#/c/441692/ | |
| 18:20:35 | mriedem | i think i addressed all of johnthetubaguy's comments | |
| 18:20:43 | mriedem | i'm not sure if there is something i need to update in there, so i haven't | |
| 18:21:02 | openstackgerrit | Peter Hamilton proposed openstack/nova-specs master: Add support for certificate validation https://review.openstack.org/357151 | |
| 18:24:26 | sdague | mriedem: have you tried to POC any of that yet? | |
| 18:24:31 | mriedem | sdague: no | |
| 18:24:49 | mriedem | i started by just trying to get devstack setup with searchlight in a ci job | |
| 18:25:17 | mriedem | https://review.openstack.org/#/c/441696/ | |
| 18:25:18 | sdague | how did that go? | |
| 18:25:21 | mriedem | not well | |
| 18:25:35 | mriedem | i didn't spend much time on it | |
| 18:25:45 | mriedem | i was hoping the searchlight devstack plugin would configure nova, but it doesn't | |
| 18:26:10 | mriedem | and now the logs are gone so i don't remember what was failing, i'll recheck it | |
| 18:26:26 | mriedem | also, searchlight doesn't have a dsvm job, so they didn't have something i could copy really | |
| 18:26:28 | sdague | so, it feels like one of those things were could define a direction here as we're going to try, but there should be a bit more POC before we commit (and especially punt on doing any of this in nova) | |
| 18:26:34 | sdague | mriedem: ah, bummer | |
| 18:26:59 | mriedem | sdague: the spec is about the general plan | |
| 18:27:12 | mriedem | just like the original resource providers spec was about the general plan, and changed a lot over time | |
| 18:27:28 | sdague | ok | |
| 18:28:08 | mriedem | i also have http://forumtopics.openstack.org/cfp/details/31 proposed | |
| 18:28:40 | mriedem | i agree poc would be ideal | |
| 18:29:15 | mriedem | chances are this won't happen in nova in pike anyway, since there are dependencies on changes in searchlight first | |
| 18:29:17 | mriedem | which don't have owners | |
| 18:29:40 | mriedem | but it's The Plan (tm) | |
| 18:32:27 | sdague | yeh, I guess I'd just soften it further from the recommended mode, given we have no idea if it's going to work. Just say, we think this is going to be a better search approach, and we want to implement an optional backend to get it into the field. | |
| 18:32:46 | sdague | I don't think this is nearly as straight forward an aggressive cut over as placement | |
| 18:32:57 | openstackgerrit | Ken'ichi Ohmichi proposed openstack/nova master: WIP: Add the specified microversion into error msg https://review.openstack.org/451986 | |
| 18:32:58 | openstackgerrit | Ken'ichi Ohmichi proposed openstack/nova master: Remove unnecessary duplicated NOTE https://review.openstack.org/452284 | |
| 18:33:25 | mriedem | i wouldn't say integrating placement has been straightforward | |
| 18:40:23 | mriedem | melwitt: so, i hate to talk about this, but before i fix the bug that's shown in https://review.openstack.org/#/c/449334/ - i'm thinking about what you said where were are committing the quota usage decrement in the api before we delete the instance, and on failure we try to rollback, but that doesn't work | |
| 18:40:52 | mriedem | i noticed something funny in the quota object | |
| 18:41:07 | melwitt | okay | |
| 18:41:10 | mriedem | on commit and rollback, if the reservation is not set, it's a noop | |
| 18:41:11 | mriedem | https://github.com/openstack/nova/blob/master/nova/objects/quotas.py#L100 | |
| 18:41:16 | mriedem | https://github.com/openstack/nova/blob/master/nova/objects/quotas.py#L111 | |
| 18:41:22 | mriedem | i was wondering why we didn't fail | |
| 18:41:38 | mriedem | but those are really invalid operations, if you're trying to commit or rollback reservations before you made the reservation, it's a bug | |
| 18:41:40 | mriedem | right? | |
| 18:41:51 | melwitt | fwiw, I think it's a no-op at the db/api level too, so I don't think it matters | |
| 18:42:06 | mriedem | i'm not sure that's a good reason though | |
| 18:42:13 | melwitt | sure | |
| 18:42:27 | mriedem | but yeah, in the db api methods, if they don't find a reservation, they don't do anything | |
| 18:42:36 | melwitt | yeah, it's not valid to do a commit/rollback without a reserve first | |
| 18:42:36 | mriedem | i was wondering why we didn't get a NotFound error on delete | |
| 18:42:55 | mriedem | https://github.com/openstack/nova/blob/master/nova/db/sqlalchemy/api.py#L3934 | |
| 18:43:11 | mriedem | https://github.com/openstack/nova/blob/master/nova/db/sqlalchemy/api.py#L3922 | |
| 18:43:13 | melwitt | I'm not sure what the reasoning is to do a no-op, if there's some case where that would be a thing | |
| 18:43:15 | mriedem | reservations are just uuids right? | |
| 18:43:29 | melwitt | reservations are records in the reservations tables | |
| 18:43:32 | melwitt | *table | |
| 18:43:36 | mriedem | yeah | |
| 18:43:36 | mriedem | with a uuid | |
| 18:43:40 | mriedem | filter(models.Reservation.uuid.in_(reservations)).\ | |
| 18:44:00 | mriedem | https://github.com/openstack/nova/blob/master/nova/db/sqlalchemy/api.py#L3866 | |
| 18:44:03 | mriedem | is on reserve | |
| 18:44:06 | mriedem | we get the list of uuids back | |
| 18:44:09 | mriedem | and store that in the Quota object | |
| 18:44:20 | artom | Can an instance have multiple fixed IPs? | |
| 18:44:36 | mriedem | so maybe i'll tinker with making commit/rollback w/o a reservation in the object a failure | |
| 18:44:42 | melwitt | artom: I think so, if it's on multiple networks | |
| 18:44:43 | mriedem | and see what falls out | |
| 18:44:49 | artom | melwitt, thanks :) | |
| 18:44:51 | mriedem | artom: yeah definitely | |
| 18:44:53 | clarkb | mriedem: artom also ipv4 + ipv6 | |
| 18:44:54 | mriedem | multinic | |
| 18:45:03 | clarkb | er melwitt artom ^ | |
| 18:45:08 | mriedem | and ipv8 | |
| 18:45:11 | mriedem | it's secret | |
| 18:45:12 | mriedem | and new | |
| 18:45:19 | mriedem | full of tomatoe | |
| 18:45:27 | artom | Are we running out of quarks to give IPv6s to? | |
| 18:45:38 | artom | IoSAP? | |
| 18:45:47 | artom | Internet of SubAtomic Particles... | |
| 18:47:05 | artom | Eesh, one bad joke and everyone shuts up? | |