Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-08-02
19:57:11 jaypipes kk
19:58:16 jaypipes dansmith: you want to keep those LOG.info() lines in remove_provider_from_instance_allocation()?
19:58:30 mriedem this deals with the various disk things https://github.com/openstack/nova/blob/master/nova/scheduler/client/report.py#L154
19:58:37 mriedem so throw that into a utility method if we need
19:58:43 jaypipes mriedem: yup, on it.
19:59:28 mriedem dansmith: you added this back in https://review.openstack.org/#/c/490085/4/nova/tests/functional/test_servers.py@1300
20:00:23 mriedem can i work on cleaning up https://review.openstack.org/#/c/490085/ and squashing in https://review.openstack.org/#/c/490159/ ?
20:00:38 mriedem this is quite a turducken we've gotten ourselves into
20:01:39 dansmith mriedem: ah yeah I was swapping things around several times when I couldn't get his patch working on master, sorry about that
20:01:45 mriedem so i can fix?
20:01:59 dansmith mriedem: jay's working on top of it, but if he's willing to rebase
20:02:00 dansmith your call
20:02:12 dansmith his call
20:02:19 mriedem what i'm changing won't impact jay probably
20:02:25 dansmith aye, just the rebase
20:02:26 mriedem so i'm going for it
20:02:27 jaypipes I can rebase no worries.
20:02:42 dansmith that was not fun
20:02:46 dansmith I need to do something more fun now
20:02:47 jaypipes dansmith: I'm going to leave those LOG.info() lines in remove_provider_from_allocations() but just pep8 em up.
20:02:51 dansmith like smash my finger in a door or something
20:02:56 mriedem i have a fun story
20:02:59 mriedem unrelated to this
20:03:01 dansmith jaypipes: ack
20:03:13 mriedem you know how horizon lets you edit a flavor?
20:03:14 openstackgerrit Sean Dague proposed openstack/nova master: add top 404 redirect https://review.openstack.org/490181
20:03:15 openstackgerrit Sean Dague proposed openstack/nova master: sort redirectmatch lines https://review.openstack.org/490182
20:03:23 dansmith mriedem: "edit" yes
20:04:13 mriedem we've got a customer that hit a fun scenario where they'd change the flavor on an instance in horizon, and it would show the old flavor even though it was deleted b/c of read_deleted='yes' from the nova db
20:04:29 mriedem then they upgraded, or something, and same scenario, but this time flavor 404 b/c no soft delete in the api db,
20:04:31 mriedem which we know about,
20:04:33 mriedem but horizon....
20:04:37 openstackgerrit Chris Dent proposed openstack/nova master: Update RT aggregate map less frequently https://review.openstack.org/489633
20:05:02 mriedem fixed in 2.47,
20:05:09 mriedem but this is like mitaka -> pike
20:05:23 mriedem good times
20:05:27 sdague mriedem: right, that was the crux of the fight around whether the original flavor id was included in the embedded flavor structure
20:05:35 sdague because of that feature in horizon
20:05:49 mriedem we should have added a cell0 for flavors :)
20:05:54 mriedem nova_flavors db
20:06:04 sdague which lets people dig themself a hole to fall in after they shoot themselves in the foot
20:06:05 mriedem when you just can't get enough dbs
20:06:21 sdague I'm telling you, db per project
20:06:27 mriedem sounds like a fun game
20:10:10 openstackgerrit Rawan Herzallah proposed openstack/nova master: Adding NVMEoF for libvirt driver https://review.openstack.org/482640
20:12:10 mriedem so i think Kevin_Zheng is going to send something to the ML asking about changing that behavior in horizon
20:12:29 dansmith meaning removing flavor editing?
20:12:37 openstack Launchpad bug 1708260 in OpenStack Compute (nova) "Sending empty allocations list on a PUT /allocations/{consumer_uuid} results in 500" [Medium,Confirmed] - Assigned to Chris Dent (cdent)
20:12:37 cdent dansmith, jaypipes, edleafe: https://bugs.launchpad.net/nova/+bug/1708260 you want to provide an opinion on whether the response should be a 400 or a lukewarm success (you asked me to do nothing, I have successfully done nothing)
20:12:46 mriedem at least disabling the ability to do it on the instance record itself,
20:13:03 mriedem like, there is a panel showing the instance and it's flavor and the ability to edit it right there on the instance record
20:13:09 mriedem which we know doesn't actually resize the instance or anythign
20:13:22 dansmith cdent: 200 vs 400 seems like something I'd have an uninformed gut opinion on, which you would immediately whip out some document to refute
20:13:26 dansmith cdent: so... no, I don't care :)
20:13:30 dansmith cdent: 500 seems wrong
20:13:55 cdent dansmith: aw dan, I’m trying to be inclusive.
20:14:06 mriedem 409
20:14:09 mriedem always 409
20:14:35 dansmith cdent: I'm being tongue-in-cheeky
20:14:44 cdent me too
20:15:09 cdent I’m going to call is a schema violation for now
20:15:35 openstackgerrit Jackie Truong proposed openstack/nova master: Add trusted_certs to Instance object https://review.openstack.org/489408
20:15:43 edleafe If the allocations body is required, 409 seems like the correct response
20:16:07 mriedem ha
20:16:10 mriedem wouldn't it be a 400?
20:16:42 cdent that is, like, so wrong
20:16:47 cdent it would be 400
20:16:54 mriedem 402?
20:16:59 mriedem i'll take your nothing and charge you for it
20:17:08 mriedem haha
20:17:28 edleafe sorry, fat finger
20:17:32 edleafe yes, 400
20:17:41 dansmith i'll hit the brakes and he'll fly right by
20:18:21 sdague 418 and call it a day :)
20:18:23 cdent so anyway, edleafe: you’ve hit crux of my query: is it required?
20:18:41 cdent one could argue that sending an empty allocations list is some kind of delete
20:18:45 cdent but that just feels icky
20:18:51 cdent so I’m going to change the schema
20:19:15 sdague cdent: 400 on that and call it a schema violation seems right to me
20:19:37 edleafe cdent: Having an allocation list is required. Populating it with allocations is not
20:19:45 sdague the point of doing that is to help the user write a better application by erroring on them when they might have typoed a thing
20:19:46 edleafe So yeah, an empty list == delete
20:19:55 dansmith mriedem: come on I get no love for the 80s movie reference?
20:20:33 sdague edleafe: a delete should be explicit, and not mistaken for a forgotten initialization
20:21:12 edleafe sdague: agree. So a schema change seems like the best solution
20:21:18 mriedem dansmith: was coding
20:21:26 mriedem i see
20:21:26 mriedem oh top gun
20:21:28 mriedem good one mav
20:22:56 cdent edleafe, sdague: it is likely that before we added the consumer_id to the object, it operated as an unintentional delete
20:23:08 dansmith by the way, I'll be taking up a collection to buy one of these for mriedem: https://chummytees.com/products/i-speak-fluent-movie-quotes-t-shirt-hoodie-tank-top
20:24:05 cfriesen I was waiting for "she's real fine my 409"
20:24:08 melwitt he can wear it when he gives summit presentations
20:25:30 mriedem cfriesen: isn't that beach boys ala 60s?
20:25:34 mriedem too old
20:25:37 openstackgerrit Chris Dent proposed openstack/nova master: [placement] Require at least one allocation when PUT https://review.openstack.org/490195
20:26:09 mriedem ha yup, 1962
20:26:19 cdent what’s your limit?
20:26:29 mriedem well i got the reference
20:26:33 mriedem so limit=01

Earlier   Later