| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2021-03-25 | |||
| 09:06:32 | bauzas | gibi: I don't know what to say | |
| 09:06:53 | bauzas | stephenfin got me a -1 because of the instance_type thing | |
| 09:07:14 | bauzas | and he worked on some DNM FUP so I could squash | |
| 09:07:32 | bauzas | I'm happy to see the first DNM be accepted by Zuul and the multinode job | |
| 09:07:49 | bauzas | but the second DNM got timeout issues on the integrated compute job | |
| 09:08:12 | bauzas | so the situation is, either we accept the patch as it is, but we will keep instance_type until 7.0 | |
| 09:08:40 | bauzas | or, we really want it and then we need to provide a new revision with the check pipeline to run again today | |
| 09:09:04 | gibi | the new revision would be the squash is it? | |
| 09:09:08 | bauzas | yup | |
| 09:09:25 | bauzas | but there is still an open question about the squashed changes | |
| 09:09:31 | bauzas | ideally, I'd squash the three | |
| 09:09:40 | bauzas | as we also need to amend the client | |
| 09:10:14 | bauzas | so I'm grateful for stephenfin's work and we're fortunate that we don't really check the argument keywords | |
| 09:10:23 | bauzas | that was one of my main concerns | |
| 09:10:53 | bauzas | but we're putting the API bump at risk while we're so close to the RC1 branching | |
| 09:11:31 | bauzas | honestly, I don't wanna rush into merging things | |
| 09:11:34 | gibi | so the server side rename is green and the client side rename is red due to timeout | |
| 09:11:44 | bauzas | -ish but yeah | |
| 09:12:05 | gibi | do we need to land the client side rename together with the 6.0 bump in W? or is it something that can be done later in X? | |
| 09:12:48 | bauzas | that's the client side | |
| 09:12:58 | bauzas | but we still have computes who use this client side, right? | |
| 09:13:34 | bauzas | a deployment with mixed Xena and Wallaby computes couldn't know about the rpc changes | |
| 09:13:44 | bauzas | at least for the W computes | |
| 09:13:48 | bauzas | but, | |
| 09:13:58 | bauzas | this is for prep_resize and resize_inst() | |
| 09:14:10 | bauzas | AFAIK, only conductors are using this rpc client | |
| 09:15:10 | bauzas | honeslty, I'm a bit afraid of rushing RPC bits based on our confidence | |
| 09:15:38 | bauzas | so, I'd say, punt this RPC bump to Xena | |
| 09:15:41 | bauzas | alas. | |
| 09:27:45 | gibi | bauzas: I guess besides the risk of late change, we also short on time to let the squash goes through the check queue today | |
| 09:27:59 | bauzas | that's my point | |
| 09:28:18 | bauzas | if we squash, we enter the odds of the check pipeline | |
| 09:28:42 | gibi | if we only merge the 6.0 as is today, then the rename needs to wait for 7.0 I guess | |
| 09:28:53 | bauzas | but I certainly understand stephenfin's point and dansmith had another argument about not supporting 5.0 | |
| 09:29:18 | bauzas | so, either we say we accept the current change, or we punt to Xena honestly | |
| 09:29:32 | bauzas | I'm not paid by any customer | |
| 09:29:39 | bauzas | so I can leave it die | |
| 09:51:17 | stephenfin | bauzas: There has been two runs of the client change. Between the two every test passed so that's good | |
| 09:51:29 | stephenfin | so let's just squash and approve? | |
| 09:51:39 | stephenfin | that gives us the day to get it merged | |
| 09:53:09 | bauzas | we need to branch RC1 today EOB | |
| 09:54:38 | stephenfin | then we'd want to get moving :) | |
| 09:56:05 | openstackgerrit | Stephen Finucane proposed openstack/nova master: Bump the Compute RPC API to version 6.0 https://review.opendev.org/c/openstack/nova/+/782935 | |
| 09:56:23 | stephenfin | like that (different change ID) | |
| 09:58:47 | stephenfin | I don't think worrying about the check queue is a good way to think about it. We're going to have the exact same issue if it fails the gate, which is quite possible based on current trends | |
| 09:59:45 | bauzas | stephenfin: then, we need two pairs of eyes | |
| 09:59:52 | stephenfin | gibi: ^ ? | |
| 09:59:55 | bauzas | gibi: can you look at the DNM ? | |
| 10:00:14 | bauzas | also, we need to also squash the rpc client | |
| 10:00:15 | gibi | I've looked at the two extra patches. | |
| 10:00:26 | gibi | I don | |
| 10:00:28 | gibi | I don | |
| 10:00:39 | gibi | I don't know how big of risk we take | |
| 10:00:57 | stephenfin | bauzas: Obviously if you want to do the squash that's okay with me. Just doing this in case it helps to move things along | |
| 10:01:11 | stephenfin | What's the risk? That it won't make RC1? | |
| 10:01:55 | gibi | that what we have today will be RC1 and it means we have ~0 days to let the rpc bump sit on master and get feedback if it breaks someting unexpected | |
| 10:02:16 | gibi | sure we have 2 weeks to provide RC2 | |
| 10:04:14 | gibi | I remember the accel_uuid RPC break that we discovered weeks later after merge | |
| 10:05:15 | gibi | but meh, whathever we merge today it will not have time to sit on master before it becomes stable/wallaby | |
| 10:05:53 | gibi | I did not see anything obviously wrong in the two extra patches but that is just me. | |
| 10:06:03 | stephenfin | I think this is an issue we've to deal with regardless of which patches go in | |
| 10:06:05 | bauzas | gibi: we can't merge it for RC2 | |
| 10:06:12 | gibi | bauzas: I know | |
| 10:06:20 | gibi | it is either today or Xena | |
| 10:06:25 | bauzas | yup | |
| 10:06:40 | bauzas | b/c we can't backport it down to stable/wallaby after today | |
| 10:07:03 | gibi | does Dan reviewed the two extra patches? | |
| 10:07:04 | bauzas | and this is not a regression | |
| 10:07:11 | bauzas | gibi: nope, I don't think | |
| 10:07:22 | stephenfin | He did yeah | |
| 10:07:30 | bauzas | we need to doublecheck the jobs we use | |
| 10:07:48 | stephenfin | he gave me feedback on fixing the client too | |
| 10:07:51 | bauzas | IIRC, grenade-multinode calls the resize API | |
| 10:08:09 | bauzas | stephenfin: again, you were fortunate we don't pass kwargs | |
| 10:08:18 | stephenfin | I know, but we don't so it's okay | |
| 10:08:37 | bauzas | because you're passing something named instance_type to something awaiting to get flavors | |
| 10:08:57 | bauzas | so maybe dansmith could have concerns with this | |
| 10:09:09 | stephenfin | the name of args doesn't matter | |
| 10:09:15 | stephenfin | it's only kwargs that matter | |
| 10:09:18 | stephenfin | standard Python | |
| 10:09:40 | stephenfin | I could call it my_terrible_terrible_awful_variable_name if I wanted to and things would be rost | |
| 10:09:42 | stephenfin | *rosy | |
| 10:09:46 | stephenfin | :) | |
| 10:10:09 | stephenfin | and the successful CI jobs prove as much | |
| 10:12:38 | stephenfin | bauzas: the grenade job runs resize tests https://zuul.opendev.org/t/openstack/build/07e5714177574a1fb513f4ae6b83bbc5/log/job-output.txt#4617 | |
| 10:12:53 | stephenfin | e.g. tempest.api.compute.servers.test_server_actions.ServerActionsTestJSON.test_resize_server_confirm | |
| 10:12:58 | bauzas | yup, was looking into it | |
| 10:13:39 | bauzas | I knew we had tempest tests but I was wondering whether we were calling them with the grenade multinode job | |
| 10:13:41 | gibi | I OK, so we have to move forward somehow. and I guess as a PTL I need to make the decision | |
| 10:13:54 | stephenfin | so we've got test coverage, the tests are passing and the code is sane (IMO at least) | |
| 10:15:03 | stephenfin | whatever risk exists with these changes also exists with the main change, so I don't see any reason to not include this if we're doing the RPC bump (which I do think we should, fwiw) | |
| 10:15:06 | stephenfin | gibi: pretty much | |
| 10:15:49 | gibi | I will take the risk of breaking things and I will use you both to fix any breakage if happens | |
| 10:15:58 | gibi | is that fair? | |
| 10:16:03 | stephenfin | wfm | |
| 10:16:31 | gibi | bauzas: ? | |
| 10:16:49 | bauzas | gibi: use your stick. | |
| 10:17:08 | bauzas | :p | |
| 10:17:17 | gibi | I have broom ;) | |
| 10:17:40 | bauzas | well, I don't ask you fifty nuances of -1s | |