| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2022-06-22 | |||
| 07:07:05 | bauzas | gibi: today is review day for me | |
| 07:07:26 | gibi | bauzas: I felt I needed to give some guide to encourage reviewers | |
| 07:10:00 | bauzas | :) | |
| 07:10:16 | bauzas | I guess we will be only a few of them | |
| 07:10:32 | gibi | bauzas: I have to respin the series due to a doc issue | |
| 07:10:49 | gibi | just I want to finish the downstream backport review first | |
| 07:10:51 | bauzas | gibi: ok, no worries, I'll start with Uggla series | |
| 07:11:06 | bauzas | after a coffee... | |
| 07:11:35 | gibi | sure | |
| 08:37:56 | bauzas | Uggla: voted on https://review.opendev.org/c/openstack/nova/+/831507/14 | |
| 08:38:25 | bauzas | Uggla: tl;dr: yeah IMHO you could squash your patch with gibi's one but I'd also want you to modify how you use the sentinel | |
| 08:40:16 | Uggla | bauzas, ok I'm going to have a look and propose an updated version ASAP. | |
| 08:40:23 | bauzas | Uggla: thanks | |
| 08:40:41 | bauzas | with gibi's patch and just the thing I ask you, I don't think this is massive change | |
| 08:40:56 | bauzas | Uggla: one thing I also wrote, please bear with logs | |
| 08:41:27 | bauzas | Uggla: the point is, logs are very well appreciated, but large public clouds tend to disagree with us adding more logs :) | |
| 08:42:01 | bauzas | in particular when you can get information from other services, like scheduler | |
| 08:42:18 | Uggla | bauzas, oh ok. | |
| 08:42:34 | bauzas | no worries, this is something not largely known | |
| 08:42:53 | bauzas | but yeah, logs rotation are costly for them | |
| 08:43:13 | bauzas | DEBUG level is a bit different tho | |
| 08:43:23 | bauzas | but with INFO, we need to be cautious | |
| 08:43:56 | bauzas | I'll then switch to your Manila spec | |
| 08:44:16 | bauzas | the sooner we have it, the earlier we could merge the os-traits change | |
| 08:46:59 | Uggla | bauzas, anyway thx for the review | |
| 08:47:12 | bauzas | np pb, apologies this was that late | |
| 08:47:42 | bauzas | Uggla: I briefly reviewed your REST change, nothing controversial appeared | |
| 08:47:48 | bauzas | but I didn't vote | |
| 08:48:51 | Uggla | bauzas, regarding the manila one, Goutham sent comments yesterday. | |
| 08:49:18 | bauzas | nice | |
| 08:51:07 | Uggla | bauzas, remaining blocking point is the "extra spec / trait" if we want to have this explicit or not. I would rather explicit one. But we need a consensus here. | |
| 08:51:59 | Uggla | I would like to discuss this point with bauzas, gibi, sean-k-mooney. | |
| 08:52:30 | gibi | I'm around | |
| 08:53:16 | Uggla | gibi, o/ | |
| 08:54:28 | gibi | o/ | |
| 08:58:29 | bauzas | Uggla: gibi: give me 30 mins so I can recharge the context by reviewing the spec :) | |
| 08:58:52 | gibi | bauzas, Uggla: do we want to wait for sean-k-mooney too? | |
| 08:59:18 | bauzas | gibi: probably, again I need to review the spec | |
| 08:59:31 | Uggla | gibi, if possible yes. | |
| 08:59:31 | gibi | sure go ahead an review | |
| 09:00:21 | bauzas | gibi: Uggla: we could do some session by 14:30 our time if needed | |
| 09:10:41 | opendevreview | Balazs Gibizer proposed openstack/nova master: Extend device_spec with resource_class and traits https://review.opendev.org/c/openstack/nova/+/846218 | |
| 09:10:42 | opendevreview | Balazs Gibizer proposed openstack/nova master: Reject mixed VF rc and trait config https://review.opendev.org/c/openstack/nova/+/846436 | |
| 09:10:42 | opendevreview | Balazs Gibizer proposed openstack/nova master: Reject PCI dependent device config https://review.opendev.org/c/openstack/nova/+/846435 | |
| 09:10:43 | opendevreview | Balazs Gibizer proposed openstack/nova master: Reject devname based device_spec config https://review.opendev.org/c/openstack/nova/+/846466 | |
| 09:10:43 | opendevreview | Balazs Gibizer proposed openstack/nova master: Ignore PCI devs with physical_network tag https://review.opendev.org/c/openstack/nova/+/846219 | |
| 09:10:44 | opendevreview | Balazs Gibizer proposed openstack/nova master: Stop if tracking is disable after it was enabled before https://review.opendev.org/c/openstack/nova/+/847009 | |
| 09:10:44 | opendevreview | Balazs Gibizer proposed openstack/nova master: Support [pci]device_spec reconfiguration https://review.opendev.org/c/openstack/nova/+/846470 | |
| 09:21:08 | bauzas | gibi: sean-k-mooney: Uggla yeah, so we have two concerns for the Manila spec | |
| 09:21:17 | bauzas | 1/ about flavor vs. image | |
| 09:21:29 | bauzas | 2/ about how to lock a share | |
| 09:21:53 | bauzas | if you want, we can discuss this by 1230UTC, ie. 14:30 for me and 13:30 for sean | |
| 09:22:31 | Uggla | bauzas, I think 1 is more explicit / implicit trait requirement. | |
| 09:22:54 | Uggla | bauzas, anyway ok for 14:30. | |
| 09:23:04 | gibi | yepp it is about asking the user to know about all technical prereqs for attaching a manila share | |
| 09:33:17 | sean-k-mooney[m] | for manila shars you should not explictly need to specify any traits | |
| 09:33:28 | sean-k-mooney[m] | you can optionally do so but it should not be a requiement | |
| 09:33:57 | sean-k-mooney[m] | we want this feature to work with existing flavors and images | |
| 09:34:22 | sean-k-mooney[m] | so documenting that you enither need file backed memory or hugepages | |
| 09:34:34 | sean-k-mooney[m] | and having the api check for the same should be sufficent | |
| 09:35:09 | sean-k-mooney[m] | operators can use traits for the former if they want to enforce that or they can use host aggreates or az to model which host has file backed memory | |
| 09:35:28 | sean-k-mooney[m] | and just tell there users to select that az | |
| 09:35:58 | sean-k-mooney[m] | so i dont think we need to modify the flavor/image for this feature | |
| 09:38:59 | sean-k-mooney[m] | bauzas: gibi https://review.opendev.org/c/openstack/nova/+/847001 now has a functional test and i made the other changes we discussed yesterday too. | |
| 09:40:17 | gibi | sean-k-mooney[m]: it is OK to me not asking for any specific from the user in image/falvor, but checking the prereqs | |
| 09:40:39 | gibi | sean-k-mooney[m]: I will check https://review.opendev.org/c/openstack/nova/+/847001 shortly | |
| 09:41:32 | Uggla | sean-k-mooney[m], maybe I'm missing something. How can you be sure the instance will be started on a host with the proper requirement without flavor/extra spec ? | |
| 09:41:58 | sean-k-mooney[m] | the curret spec if i recall say we will check in the api if the instance has hugepages or is on a host with file backed memory adn reject the request if that is not the case | |
| 09:42:50 | Uggla | yes | |
| 09:42:53 | sean-k-mooney[m] | if the falvor request hugepages it will be on a valid host or if the user select an az that is mapped to file backed memroy | |
| 09:43:11 | sean-k-mooney[m] | so we dont need anythign explict in the flavor for file backed mory | |
| 09:43:22 | sean-k-mooney[m] | we just need to document that requirement | |
| 09:43:32 | sean-k-mooney[m] | they can also add the trait if they like | |
| 09:43:38 | sean-k-mooney[m] | but they dont need too | |
| 09:44:06 | sean-k-mooney[m] | its the same for vhost-user networking with ovs-dpdk | |
| 09:44:24 | sean-k-mooney[m] | we just document that you need file backed memory or hugepages | |
| 09:44:49 | sean-k-mooney[m] | we dont require you to use traits but you could use a cutom trait for that if you wanted too | |
| 09:45:40 | sean-k-mooney[m] | most operators are not going to want to resize all there workload to be able to start using this feature so its imporant that we dont force that when its not required. | |
| 09:45:49 | sean-k-mooney[m] | does that make sense? | |
| 09:49:45 | sean-k-mooney[m] | by the way are we accpeting translations in tree again? | |
| 09:50:58 | sean-k-mooney[m] | i just noticed https://review.opendev.org/c/openstack/nova/+/846876 | |
| 09:51:14 | sean-k-mooney[m] | but i would kind of prefer to not have those in tree | |
| 09:52:18 | Uggla | hum I think so. I'm just wonder how to check for trait at the Rest API level. (atm looking at the code for an example) | |
| 09:52:45 | sean-k-mooney[m] | you can just call placment to do a traits list on the host | |
| 09:53:51 | bauzas | sean-k-mooney: yes | |
| 09:53:56 | bauzas | +2/+Wd for me | |
| 09:53:58 | sean-k-mooney[m] | hum i guess we do get thos octaionally https://github.com/openstack/nova/commits/master/nova/locale | |
| 09:54:27 | gibi | Uggla: nova.scheduler.client.report.SchedulerReportClient.get_provider_traits | |
| 09:55:44 | bauzas | sean-k-mooney: we only had a very few translations patches since 2020 | |
| 09:55:45 | bauzas | https://docs.openstack.org/i18n/latest/reviewing-translation-import.html#reviewing | |
| 09:56:31 | sean-k-mooney[m] | yep thats why i tought we had stop doing this | |
| 09:57:51 | sean-k-mooney[m] | its fine we can proably merge it but if we are not doing a review of the content ectra it would be nicer if we just had this in a sperate repo that the i18n team could mange | |
| 09:58:07 | sean-k-mooney[m] | i guess if its that infrequent its fine | |
| 09:58:47 | sean-k-mooney[m] | if we start getting a lot of patches this way we should look at a different solution then we have currently. | |
| 09:58:49 | bauzas | sean-k-mooney: the problem is with the i18n team | |
| 09:59:00 | bauzas | sean-k-mooney: no, we won't have a lot of patches | |
| 09:59:23 | bauzas | sean-k-mooney: even if they have a lot of i18n contributors, we will only have one change | |
| 09:59:38 | bauzas | as it's an automatic import from Zanata | |
| 09:59:40 | sean-k-mooney[m] | i dont like haveing a seperate review workflow for stuff or not being able to review the change and just merging it | |
| 10:00:15 | sean-k-mooney[m] | so if that the workflwo we are going to have it think it would be better to have it entirly out of tree | |