| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-03-02 | |||
| 21:51:10 | mriedem | dansmith: that's what it looks like, the tags extensions can fail with an InstanceNotFound | |
| 21:51:15 | sdague | dansmith: I'm going to have to reload this bit back into my head | |
| 21:51:17 | mriedem | but it looks like the instance lookup is hidden in the tag object | |
| 21:51:43 | dansmith | mriedem: right I just want to make sure that's the only one, because I don't seem to make it into that method *at all* | |
| 21:51:51 | mriedem | https://github.com/openstack/nova/blob/master/nova/objects/tag.py#L51 | |
| 21:51:51 | mriedem | oh | |
| 21:51:52 | dansmith | and I'm wondering if there is some other glue that is missing the cell lookup | |
| 21:52:01 | mriedem | dansmith: microversoin? | |
| 21:52:13 | dansmith | mriedem: it's tempest | |
| 21:52:21 | mriedem | still | |
| 21:52:27 | mriedem | is it passing 2.1? | |
| 21:52:56 | mriedem | i think you need at least 2.26 | |
| 21:52:57 | dansmith | I dunno, but the test seems to work without cells. I just tried from nova client and I get the same deal | |
| 21:53:17 | mriedem | novaclient CLI will request the latest microversion available that it understands | |
| 21:53:19 | mriedem | tempest won't | |
| 21:53:35 | dansmith | PUT http://192.168.201.194:8774/v2.1/servers/0d97611d-f1d6-4146-858d-9387baf36558/tags/tagme -H "OpenStack-API-Version: compute 2.41" | |
| 21:53:36 | mriedem | i don't think anyway | |
| 21:53:37 | sdague | mriedem: the tests should be configured correctly | |
| 21:53:43 | cfriesen | related to my earlier DB question....is there a reason why "instance_actions" rows don't get deleted when the instance does? | |
| 21:53:58 | mriedem | cfriesen: have you bothered to look at the code? | |
| 21:54:45 | cfriesen | mriedem: I'm looking at instance_destroy() in the DB API code | |
| 21:55:05 | mriedem | it doesn't appear that they do there | |
| 21:55:17 | mriedem | i believe this came up in the ML in newton | |
| 21:55:30 | cfriesen | okay, I can check the archives | |
| 21:55:37 | sdague | dansmith: I'm pretty sure nothing else is validating the server id until you get into server_tags.py | |
| 21:55:43 | cfriesen | I didn't see a bug for it | |
| 21:55:48 | dansmith | sdague: okay | |
| 21:55:55 | sdague | but, it's end of day, so that's just my recollection | |
| 21:56:06 | dansmith | sdague: http://pastebin.com/bicZYZVt | |
| 21:56:20 | dansmith | sdague: with that, when I add a tag from the client, I get _none_ of my log lines in the log | |
| 21:56:36 | dansmith | well, when I _try_ to add a tag.. it fails with 404 | |
| 21:57:48 | dansmith | anyway, just want to make sure I'm not missing something obvious | |
| 21:57:58 | mriedem | cfriesen: https://github.com/openstack/nova/blob/master/nova/db/sqlalchemy/api.py#L6468 | |
| 21:58:07 | mriedem | they are handled differently for archive at least | |
| 21:58:11 | mriedem | because they aren't soft deleted | |
| 21:59:27 | mriedem | cfriesen: http://lists.openstack.org/pipermail/openstack-dev/2015-November/079701.html | |
| 22:00:13 | dansmith | mriedem: just checking, but... api_version('2.26') means "2.26 or later" right? just making sure we don't have another single-version thing | |
| 22:00:20 | dansmith | that wouldn't explain tempest not working though | |
| 22:00:37 | cfriesen | mriedem: I was just looking at the bug related to that email thread | |
| 22:00:48 | mriedem | cfriesen: http://lists.openstack.org/pipermail/openstack-dev/2015-November/079778.html | |
| 22:00:50 | mriedem | ^ is the summary | |
| 22:01:13 | sdague | dansmith: https://github.com/openstack/nova/blob/37cd9a961b065a07352b49ee72394cb210d8838b/nova/api/openstack/wsgi.py#L907-L910 | |
| 22:01:22 | mriedem | dansmith: depends on where the code is i guess | |
| 22:01:41 | dansmith | sdague: okay yeah | |
| 22:02:35 | sdague | dansmith: anyway, if this is still a thing tomorrow, I can help debug. But I'm pretty sure we end up just building the routes through route mapper and we get straight to these functions | |
| 22:02:43 | mriedem | cfriesen: i also remember this coming out of that discussion https://docs.openstack.org/developer/nova/api_microversion_history.html#id19 | |
| 22:02:50 | dansmith | sdague: okay me too, I just don't know how else to explain what I'm seeing | |
| 22:02:55 | dansmith | sdague: but yeah thanks | |
| 22:02:57 | sdague | the only question is if there is a very early 404 in wsgi | |
| 22:03:00 | dansmith | right | |
| 22:03:14 | cfriesen | mriedem: right, I saw that part. | |
| 22:03:50 | sdague | https://github.com/openstack/nova/blob/37cd9a961b065a07352b49ee72394cb210d8838b/nova/api/openstack/wsgi.py#L601 | |
| 22:03:59 | sdague | but that should show up in nova-api logs as a fault | |
| 22:04:08 | openstackgerrit | Sujitha proposed openstack/nova master: Add description to policies in admin_actions. https://review.openstack.org/440829 | |
| 22:04:48 | dansmith | oooooo | |
| 22:05:16 | dansmith | no, no. | |
| 22:05:17 | cfriesen | mriedem: that summary is useful, thanks. it might be nice if it was in the code as a comment or something. | |
| 22:05:18 | dansmith | dammit | |
| 22:05:44 | sdague | ok, you have to share before I run away | |
| 22:05:48 | dansmith | there is no instance lookup in the path of the PUT | |
| 22:05:48 | sdague | because I want to know :) | |
| 22:06:02 | dansmith | I thought there was, but it's for the server get the client does before the put of the tag | |
| 22:06:05 | dansmith | after the get returns, | |
| 22:06:13 | sdague | ah | |
| 22:06:14 | dansmith | no instance get at all | |
| 22:06:19 | dansmith | so...... | |
| 22:06:24 | sdague | right, so you are way off in 404 land | |
| 22:06:35 | dansmith | I don't like 404 land | |
| 22:08:21 | openstackgerrit | Sujitha proposed openstack/nova master: Add description to policies in admin_actions. https://review.openstack.org/440829 | |
| 22:09:46 | mriedem | dims: v | |
| 22:09:47 | mriedem | https://review.openstack.org/440832 | |
| 22:10:05 | mriedem | cfriesen: i have to go back and read the summary myself | |
| 22:10:50 | dansmith | sdague: omg I am so awesaome | |
| 22:11:36 | dansmith | sdague: RROR stevedore.extension [req-3b12728d-ad52-44bd-b79d-ba66275e1d1d None None] Could not load 'server_tags': cannot import name nova_context | |
| 22:12:35 | mriedem | cfriesen: ok so we fixed the 2 things in that thread we said we would, | |
| 22:12:44 | mriedem | which is the microversion to list instance actions for deleted instances, | |
| 22:12:53 | mriedem | and hard delete instance actions when archiving an instance | |
| 22:13:14 | mriedem | cfriesen: if you wanted to push a change to add a comment about why we don't soft delete instance actions when deleting an instance, in the instance_destroy db api method, i'd +2 that | |
| 22:13:28 | cfriesen | mriedem: okay | |
| 22:22:21 | openstackgerrit | Anusha Unnam proposed openstack/nova master: Add description to policies to availability_zone https://review.openstack.org/440836 | |
| 22:25:30 | cfriesen | one final DB question (for today, anyway). What's the story around console_auth_tokens added in Newton? Is something going to delete them? | |
| 22:25:57 | dansmith | my jar runneth over | |
| 22:28:58 | mriedem | cfriesen: talk to melwitt about the spec for that | |
| 22:29:23 | cfriesen | mriedem: I'm looking at https://blueprints.launchpad.net/nova/+spec/convert-consoles-to-objects ... looks like it got only partially implemented? | |
| 22:29:31 | mriedem | cfriesen: yes | |
| 22:29:39 | mriedem | melwitt is going to re-propose the spec for pike | |
| 22:30:25 | cfriesen | okay, cool. so in Newton the table is there but unused? | |
| 22:31:24 | melwitt | looks like it. you can see the patches that merged vs ones that ended up abandoned here https://review.openstack.org/#/q/topic:bp/convert-consoles-to-objects,n,z | |
| 22:32:18 | melwitt | so my plan is to repropose the spec and then resurrect those old patches and fix them up if needed to finish the work | |
| 22:33:12 | cfriesen | melwitt: thanks. I was looking for users of console_auth_token_create() but didn't see any. | |
| 22:33:31 | melwitt | yeah | |
| 22:36:05 | cfriesen | melwitt: okay, looks like https://review.openstack.org/#/c/325381/ introduces a periodic task to clean up expired tokens. seems maybe a bit odd to drive it from the compute nodes though. | |
| 22:38:36 | melwitt | cfriesen: yeah, although that works out for cells, since the tokens will be kept in cell databases, and only n-cpu and n-cond run in a cell | |
| 22:44:33 | jaypipes | dansmith: hey, splitting that big patch is prolly a good idea at this point, yeah. I'll do it when I get to my hotel this evening (currently sitting in SeaTac airport...) | |
| 22:44:48 | dansmith | jaypipes: okay I think it'll be easy | |
| 22:45:51 | cfriesen | just noticed something odd. at https://github.com/openstack/nova/blob/master/nova/db/sqlalchemy/api.py#L1879 we soft-delete migrations when the instance is deleted, but then at https://github.com/openstack/nova/blob/master/nova/db/sqlalchemy/api.py#L6474 we special-case the migrations table like we do for instance_actions. Is that no longer needed? | |
| 22:47:30 | jaypipes | dansmith: ya, shouldn't be a big deal. | |
| 22:47:35 | mriedem1 | cfriesen: that's for dealing with migrations that were not soft deleted | |
| 22:47:51 | mriedem1 | so you'd have migrations in the db that weren't soft deleted and those would break the archive | |
| 22:48:07 | mriedem1 | anything after we started soft deleting migrations when the instance was deleted would be fine | |