| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-05-25 | |||
| 14:46:24 | cdent | edleafe: ping me when you start, I'd like to ride along please | |
| 14:46:41 | edleafe | cdent: roger that | |
| 14:47:32 | sdague | mriedem: the context bits look good to you now? | |
| 14:47:59 | sdague | given that we're blocking a constraint update on that | |
| 14:48:10 | mriedem | sdague: +1 until dansmith can take a look | |
| 14:48:19 | sdague | ok, cool | |
| 14:48:38 | dansmith | mriedem: sdague: I got a call starting in a few and then I'll look | |
| 14:48:45 | sdague | dansmith: sounds good | |
| 14:48:48 | sdague | thank you | |
| 14:49:02 | sdague | mriedem: I have an orthoginal (but related question) | |
| 14:49:27 | sdague | for nova to be a good player in global request id, and accept them, we need https://review.openstack.org/#/c/467998/ | |
| 14:49:33 | alex_xu | mriedem: yea, i think so. and interesting bug, add it into review list | |
| 14:49:39 | sdague | which is going to cause an extra http header to emit | |
| 14:49:50 | sdague | microversion? | |
| 14:49:57 | mriedem | alex_xu: i'm fixing the test in that one | |
| 14:50:10 | sdague | alex_xu might also want to weigh in there | |
| 14:50:46 | mriedem | sdague: i tend to say yes, otherwise how do you know if the cloud has the header or not? | |
| 14:50:48 | mriedem | however, | |
| 14:51:03 | mriedem | do we hook in something that removes the header for requests < new microversion? | |
| 14:51:10 | sdague | nope | |
| 14:51:22 | mriedem | otherwise the new microversion is only signaling | |
| 14:51:27 | sdague | correct | |
| 14:51:42 | sdague | also, it depends on code version on the server and if they follow our upper constraints | |
| 14:52:04 | sdague | it's definitely a little messy, and mostly annoying we didn't do this a long time ago | |
| 14:52:16 | mriedem | sdague: we can't require a new minimum version of oslo.middleware? | |
| 14:52:39 | sdague | maybe, I can look at the force on the requirements side | |
| 14:52:48 | sdague | we could land it with the requirements force | |
| 14:55:16 | mriedem | you want oslo.middleware>=3.27.0 it looks like | |
| 14:55:20 | mriedem | for https://github.com/openstack/oslo.middleware/commit/0e3dbba0d9d373fbfa9e28897cd33377e2491d32 | |
| 14:56:32 | sdague | yep | |
| 14:56:40 | alex_xu | emm...any special reason for not having a microversion? | |
| 14:56:51 | sdague | https://review.openstack.org/#/c/468057/ | |
| 14:57:46 | sdague | mriedem: ok, I'll build on top of that in my spider of crazy depends on | |
| 15:01:28 | edleafe | jaypipes: free? | |
| 15:01:37 | jaypipes | edleafe: sorry, still on call. | |
| 15:01:46 | jaypipes | edleafe: will ping you asap. | |
| 15:02:30 | edleafe | jaypipes: kewl | |
| 15:05:21 | openstackgerrit | Sean Dague proposed openstack/nova master: Migrate to oslo request_id middleware - mv 2.46 https://review.openstack.org/467998 | |
| 15:06:44 | mriedem | sdague: you'll throw a shoe, but we should have a bp and spec for that microversion | |
| 15:06:55 | mriedem | i imagine it will be small | |
| 15:07:36 | sdague | mriedem: specless blueprint? | |
| 15:07:39 | cdent | I should warn you, I do no the other one, so if you like that one, you're gonna hobble for a long time | |
| 15:07:49 | mriedem | sdague: all api changes need a spec | |
| 15:07:53 | mriedem | thems the rules | |
| 15:09:17 | mriedem | smcginnis: so what kind of fun dual maintenance path is it going to be for os-brick in both python and go? | |
| 15:09:31 | sdague | mriedem: ok, let me throw together something quick | |
| 15:10:51 | mriedem | jaypipes: isn't this vestigial now? https://github.com/openstack/nova/blob/master/nova/db/sqlalchemy/api.py#L800 | |
| 15:11:49 | mriedem | added in https://github.com/openstack/nova/commit/8941b45bc58a833261bb09fc7eb8fb5c765e154a | |
| 15:12:18 | mriedem | wondering b/c of https://github.com/openstack/nova/commit/ea4754ba21ea08eb367d62bc46b5eba3777c119c | |
| 15:12:23 | jaypipes | mriedem: yeah. the comment can be removed. | |
| 15:20:49 | smcginnis | mriedem: Eh, it's inevitable. | |
| 15:24:28 | mriedem | hmm, i've got questions about compute nodes and services records | |
| 15:24:34 | mriedem | b/c something isn't adding up | |
| 15:25:44 | mriedem | deleting a nova-compute service record doesn't delete the compute_nodes record, | |
| 15:26:07 | mriedem | but if you restart the nova-compute process, it's going to create a new services table record for the same host/binary/topic | |
| 15:26:15 | mriedem | the compute_nodes table entry has a service_id record in it | |
| 15:26:33 | mriedem | the resource tracker isn't going to create a new compute_nodes record because one already exists for the given host and nodename | |
| 15:26:43 | mriedem | but, do we ever update the compute_nodes.service_id record anywhere? | |
| 15:27:46 | openstackgerrit | Sean Dague proposed openstack/nova-specs master: oslo.middleware request_id spec https://review.openstack.org/468066 | |
| 15:27:52 | sdague | mriedem: ok, spec ^^^ | |
| 15:28:17 | mriedem | ok | |
| 15:30:19 | mriedem | jaypipes: do we actually not tie the compute nodes record to the nova-compute service anywhere? | |
| 15:31:31 | mriedem | ah looks like it's not, by design | |
| 15:31:34 | mriedem | since it's all deprecated | |
| 15:35:53 | mriedem | oh snap | |
| 15:36:06 | mriedem | alex_xu: we do soft delete the compute node record when the service is deleted, | |
| 15:36:09 | mriedem | if they are tied together, | |
| 15:36:15 | mriedem | which happens for very old compute nodes records | |
| 15:36:52 | mriedem | https://github.com/openstack/nova/blob/65402e1aeb770d077620f2935a0cb305474aa7a1/nova/db/sqlalchemy/api.py#L461 | |
| 15:37:15 | jaypipes | mriedem: we removed the direct linkage between the two on purpose | |
| 15:37:27 | mriedem | https://github.com/openstack/nova/blob/65402e1aeb770d077620f2935a0cb305474aa7a1/nova/db/sqlalchemy/api.py#L461 | |
| 15:37:28 | mriedem | yeah | |
| 15:37:37 | mriedem | the db api tests are still doing the very old style linkage | |
| 15:37:47 | mriedem | so testing using those for a modern setup is really confusing | |
| 15:46:15 | jaypipes | edleafe: k, sorry about that. ready now. | |
| 15:46:16 | jaypipes | hangout? | |
| 15:51:20 | mriedem | sdague: -1 is specifically for expanding on this https://review.openstack.org/#/c/468066/1/specs/pike/approved/oslo-middleware-request-id.rst@48 | |
| 15:51:23 | mriedem | otherwise +2 | |
| 15:52:01 | sdague | mriedem: so, it's not sufficient that the patch already has that in it? | |
| 15:52:12 | mriedem | the spec is a doc | |
| 15:52:22 | sdague | sure | |
| 15:52:24 | mriedem | published as a doc, people reading those later shouldn't have to dig through lp to find the patch | |
| 15:52:55 | edleafe | jaypipes: cdent: so you suggested passing the instance UUIDs in the reqspec | |
| 15:53:20 | edleafe | That was my original approach, but dansmith preferred the RPC parameter approach, so I changed it | |
| 15:53:38 | edleafe | jaypipes: see https://review.openstack.org/#/c/463623/ | |
| 15:53:45 | jaypipes | edleafe: right. | |
| 15:54:16 | jaypipes | edleafe: lemme walk through in my head whether the recent "continuing the automatic reschedule behaviour" thing affects my decision on that... | |
| 15:54:18 | cdent | edleafe: I don't really have a strong opinion. my natural inclination is parameters, but that's not been seasoned by years of nova-based RPC | |
| 15:55:07 | edleafe | jaypipes: cdent: my preference was reqspec (which is why I did it that way first), since it already had the pretty much useless 'instance_uuid' attribute | |
| 15:55:08 | openstackgerrit | Sean Dague proposed openstack/nova-specs master: oslo.middleware request_id spec https://review.openstack.org/468066 | |
| 15:55:13 | jaypipes | dansmith: can you, in one sentence, explain your preference to change the RPC interface vs. add a field to the reqspec object? | |
| 15:55:21 | sdague | mriedem: done | |
| 15:55:35 | jaypipes | dansmith: I want to make sure I understand you completely. | |
| 15:56:00 | dansmith | jaypipes: we don't need to store the uuids in the database, right? | |
| 15:56:22 | dansmith | we've had this conversation a couple times | |
| 15:56:44 | edleafe | dansmith: but there is already the 'instance_uuid', which we don't store either | |
| 15:57:46 | dansmith | edleafe: um, what? | |
| 15:57:53 | dansmith | already instance_uuid in the reqspec? | |
| 15:58:00 | edleafe | yup | |
| 15:58:33 | edleafe | dansmith: https://review.openstack.org/#/c/463623/5/nova/objects/request_spec.py@77 | |