Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-03-23
11:04:58 johnthetubaguy mdbooth: I couldn't think when we would fetch by attachment_id, I think I am missing something
11:05:56 mdbooth johnthetubaguy: It just seems inevitable to me if we've got apis creating by attachment
11:06:11 mdbooth However, my point was really higher level than that
11:06:30 mdbooth Which is that I don't think it's worth adding a db schema change for code we haven't seen yet
11:06:40 johnthetubaguy mdbooth: so if the index didn't have a cost, I would totally be up for adding it anyways
11:06:53 mdbooth I don't see a problem merging the code before the user
11:07:07 mdbooth I just think it's premature to merge it when we haven't even seen the user
11:07:26 mdbooth There's no reason we can't merge it quickly later
11:07:44 mdbooth Merging it early just risks potentially needing another db migration
11:08:03 mdbooth And it has (a small) cost when we're not using it
11:08:11 johnthetubaguy thats totally fair, no harm in waiting for that
11:08:48 mdbooth I was really just -1 because there's no follow up. The patch may be good.
11:09:00 johnthetubaguy I think we pass instance_uuid and volume_uuid to look up a BDM, then we check if the attachment_id is present or not, and use the new flow or not, but it would be good to see that code first
11:09:44 johnthetubaguy mdbooth: yeah, thats totally fair, I like your idea of waiting for the follow, it should answer our questions for us
11:09:58 openstackgerrit Chris Dent proposed openstack/nova master: [placement] Idempotent PUT /resource_classes/{name} https://review.openstack.org/448791
11:10:12 johnthetubaguy mdbooth: I think I managed to miss that bit of your comment, my bad
11:32:26 openstackgerrit Sylvain Bauza proposed openstack/nova master: WIP/DNM: Destroy the ReqSpec object when deleting the instance https://review.openstack.org/391060
11:33:28 openstackgerrit Sylvain Bauza proposed openstack/nova master: Destroy the ReqSpec object when deleting the instance https://review.openstack.org/391060
11:43:28 tojuvone johnthetubaguy, Hi, do you have a minute
11:47:51 tojuvone johnthetubaguy, Just wondering new host_maintenance_events table vs. using services table
11:49:00 tojuvone johnthetubaguy, services table already exposed to servers that suits me, but event would need duplicate on each service
11:54:08 tojuvone brb in an hour or get hit by major traffic jam
12:02:17 openstackgerrit Maxim Nestratov proposed openstack/nova master: WIP: honor disk_format=ploop image metadata property https://review.openstack.org/429745
12:03:56 mnestratov mdbooth: hi, could you please take a look at ^ ^ and say what you think now?
12:10:38 openstackgerrit Mikhail Feoktistov proposed openstack/nova master: Remove fstype param from ploop init https://review.openstack.org/444970
12:18:12 openstackgerrit Alexander Baranin proposed openstack/nova master: Add allocation ratios to hypervisor api. https://review.openstack.org/449107
12:19:45 sbezverk_ johnthetubaguy: ping
12:22:03 openstackgerrit Maxim Nestratov proposed openstack/nova master: tests: fix uefi testcases https://review.openstack.org/449109
12:22:33 johnthetubaguy tojuvone: sbezverk_: sorry, on a call, about to grab lunch, I should be around in 30 mins or so if thats OK
12:26:37 sbezverk_ johnthetubaguy: sure thing, bon apetit!!
12:39:23 openstackgerrit Chris Dent proposed openstack/nova master: [placement] Don't use floats in microversion handling https://review.openstack.org/449119
12:40:19 cdent johnthetubaguy, bauzas, jaypipes : that's ^ probably reasonably important enough to merge soonish
12:44:43 sbezverk_ cdent: do you have a second for qq?
12:45:07 cdent sbezverk_: I can try
12:46:14 sbezverk_ cdent: in ocata, the doc sais that we need to have two cells cell0 and cell1, but on the other hand I am running ocata with just cell0 and everything is working. SO there is an inconsistency. Could you please clarify?
12:46:15 tojuvone johnthetubaguy, No problem :)
12:46:37 sdague cdent: could you put a test that tries to set the decorator with a float to ensure that explodes really early?
12:46:44 cdent sdague: sure
12:46:55 sdague because I would be concerned by someone doing that accidentally
12:47:07 sbezverk_ cdent: is it really mandatory to have both cells?
12:47:24 Dinesh_Bhor Hi all, can someone take a look at a small patch? https://review.openstack.org/#/c/445782/ It's already having a +2.
12:47:34 cdent sbezverk_: As I understand it, yes, so something is likely not what it appears, but that's an area I'm not very familiar with
12:48:02 sbezverk_ cdent: could you suggest local cell expert ;) ??
12:48:59 sbezverk_ cdent: cause as I said I have a test bed functioning normal with single cell0.
12:49:18 cdent sbezverk_: dansmith is probably the final word there and melwitt as well but I think neither of them is around yet
12:49:48 EmilienM jroll: fyi, the ironic/nova thing is fixed for us. Thanks again for your patch!
12:49:55 cdent sbezverk_: are you sure you haven't named the thing which would be cell1 'cell0'? I believe you should have three database tables?
12:49:58 EmilienM dansmith, bauzas ^
12:50:20 bauzas EmilienM: coolio
12:50:41 sbezverk_ cdent: positive, I wrote a job and it creates a single cell0
12:51:15 sbezverk_ cdent: so I have now 3 databases only nova, nova-api and cell0
12:51:35 sbezverk_ cdent: I will follow up with these folks to confirm, thank you very much for your help
12:52:02 cdent sbezverk_: I think what you've got there is correct
12:52:25 cdent but like I said, I've not done much poking in this area, and I don't currently have a running system
12:52:56 cdent sbezverk_: actually, i do have database tables though and I have 'nova', 'nova_api' and 'nova_cell0'. only
12:53:13 cdent the table named 'nova' is cell1
12:54:07 sbezverk_ cdent: ahhh, so nova became kind of cell1, got it
12:54:40 bauzas cdent: https://review.openstack.org/#/c/449119/1 had sdague's comment that I agree with
12:54:50 cdent bauzas: yeah, just working on that now, thanks.
12:54:51 bauzas just add this and I'm +2
12:55:09 bauzas cdent: FWIW, thanks for fixing that
13:02:10 jroll EmilienM: \o/
13:06:26 cdent sdague, bauzas: If you pass in a float on that thing it will raise an AtributeError, at import time. Is that sufficient, or would you prefer that exception(s) raised from parse_version_string be more custom? I tend to prefer built in exceptions for this kind of thing.
13:07:16 bauzas cdent: I think sdague asked for an unittest verifying that behaviour
13:07:26 cdent yes, i've written the test
13:07:31 bauzas not what happens if someone uses that deco incorrectly
13:07:58 bauzas cdent: tbc, the former is enough from my PoV
13:08:20 bauzas if we have a docstring that explains the deco is awaiting for a string, and then the developer is using it incorrectly, then meh
13:09:08 bauzas it's like not following a manual when using a chainsaw
13:11:28 sdague cdent: as long as that would fail a commit I'm good with it
13:11:46 cdent sdague: yeah, code will boom
13:11:53 sdague cdent: wfm
13:13:31 cdent sdague: yeah: AttributeError: 'float' object has no attribute 'split'\nThe test run didn't actually run any tests"
13:13:48 cdent (that's on the functional tests)
13:13:53 sdague cool
13:22:34 openstackgerrit Chris Dent proposed openstack/nova master: [placement] Don't use floats in microversion handling https://review.openstack.org/449119
13:27:23 cdent sdague, bauzas: that ^ should do it
13:29:32 sdague cdent: frickler has a comment on the last patch
13:31:29 bauzas sdague: frickler missed the transformation into a tuple
13:32:35 cdent yeah, it's a tuple by then
13:33:05 sdague so, is there a reason not to move APIVersionRequest into the common library and reuse all of that?
13:33:59 cdent sdague: no reason other than time
13:34:56 cdent sdague: I also thought that writing something fresh would allow for some compare and contrast so we could have a mix of both in whatever the common thing was
13:35:08 cdent there are some differences in approach
13:35:12 cdent (nothing major)
13:35:23 sdague yeh, it feels like mostly we're just rediscovering all the bugs again :)
13:35:27 cdent heh
13:35:48 sdague I honestly remember this exact issue when we hit 2.10
13:36:03 efried lifeless Yes, https://github.com/testing-cabal/mock/issues/398 is me.
13:36:47 efried mriedem I've got a doc appt during the nova meeting today. I left my update in the agenda.
13:38:11 cdent sdague: sure. the other two issues were: didn't want to reuse nova code because I thought we'd be moving soon; and when I first did this I couldn't seem to untangle the way in which nova was doing version_handler from other parts of the wsgi stack.
13:38:30 cdent In other words I found the code too hard to read, so wrote some more. Which is sad but true.
13:39:47 cdent (it's the stuff in Controller that bit me)
13:44:25 mriedem efried: ok
13:46:16 sdague cdent: ok, but you've effectively replicated the APIVersionRequest object as a tuple now. It feels like that's going to be a pretty direct transition
13:46:37 openstackgerrit Eric Fried proposed openstack/nova master: PowerVM Driver: spawn/delete #1: no-ops https://review.openstack.org/438119
13:46:47 openstackgerrit Hironori Shiina proposed openstack/nova-specs master: Ironic: Cold migration support https://review.openstack.org/449155
13:47:31 cdent sdague; that part is not the interesting/useful/verbose part of the handling though. It is the users of the Version concept, and those parts between placement and nova are much different (simply because the framework is too)
13:47:44 cdent I agree with you that there's some duplication, and that's unfortunate, but it's where we are right now
13:48:12 cdent and I'm not too het up about repeating bug fixes, it is good for learning, for refactoring, and refreshing, especially when they are caught before they leak

Earlier   Later