Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2018-04-04
18:35:32 dansmith you need to add this as a spec_obj field, and yes, I would expect to see at least a test that uses the old dict-y spec field with the rest of the code
18:36:07 sean-k-mooney im going to get rid of the ** but im just wonding if i should get rid of all uses of spec=[{}] and move them to spec_objs=[objects]
18:36:10 dansmith I would expect to make all the existing tests use the object, and add one more new test that just makes sure something basic still works if you only have the dict field set
18:36:17 cdent sean-k-mooney, jaypipes (and dansmith and cfriesen_ you probably know this too): On jay's spec about cpu resource tracking ( https://review.openstack.org/#/c/555081/ ) the existing situation is that any given host can only be shared or dedicated, but not mixed? Is that just an OpenStack/Nova limitation, or is it also a limitation in libvirt/KVM?
18:36:34 dansmith sean-k-mooney: yeah, except for the new test to validate the old dict path
18:36:51 dansmith cdent: just nova AFAIK
18:37:24 sean-k-mooney dansmith: ok ill convert them all to use objects like i had started and then add in test just for the old path thanks.
18:37:32 dansmith sean-k-mooney: yup
18:37:38 cdent so if we get nova sorted (via jay's ideas) we'll need to cook some code in the virtdriver, but not deeper in the stack? cool.
18:38:30 sean-k-mooney cdent: am nova will happliy allow a mixed host but it is not preventing the shared cpus from stomping over the pinned ones today. so we use host aggregates to seperate them
18:39:02 cdent sean-k-mooney: ah, okah, thanks
18:39:36 dansmith right, thatg
18:41:11 melwitt bhagyashris: I have a question about your spec on deleteBackups. my assumption has been that you have users that are currently using rotation '0' to purge backups as a side effect. is that correct?
18:41:59 melwitt *using rotation '0' to the createBackup API
19:07:10 mriedem anyone know the tl;dr on the 'a plan to stop syncing requirements into projects' thread?
19:07:20 mriedem sounds like requirements.txt changes are now all done manually?
19:08:48 dansmith from the oslo room ptg,
19:09:14 dansmith I think they were saying that they would only sync (or manually sync) requirements when the minimum was being raised past what the project has listed
19:09:20 dansmith instead of just always syncing always
19:09:45 sean-k-mooney mriedem: i get why we might want to do this to prevent raising minimums higher then needed but i think we are also losing that auto blacklisting of specific versions e.g. x>1.0,x!=1.1
19:09:48 dansmith so that things like swift can continue supporting the older versions of some libraries they have because of workarounds or lack of actual requirements for the newer one
19:10:23 mriedem i just see that it's throwing people off e.g. https://review.openstack.org/#/c/554679/
19:11:36 mriedem and https://review.openstack.org/#/c/558254/
19:11:37 dansmith just yet another change in process
19:11:42 mriedem which is actually < than what's in g-r
19:11:44 openstackgerrit Tyler Blakeslee proposed openstack/nova master: Add __repr__ for NovaException https://review.openstack.org/555812
19:11:52 sean-k-mooney oh so in the scale io case they had expect the version bump but the proposal bot is not making it
19:12:05 openstackgerrit Merged openstack/nova master: [placement] api-ref: Fix parameters https://review.openstack.org/558142
19:12:18 mriedem i'm not sure why we'd even have g-r then
19:12:50 sean-k-mooney mriedem: i gues g-r keeps the minium for all pojects and nova requirments.txt has minium for nova
19:13:28 cdent mriedem: upper constraints remains the one true thing, even without a globally sync requirements.txt, maybe?
19:13:40 mriedem which means we could bump our minimum required version in nova for a library, but not care about bumping it in g-r
19:13:57 sean-k-mooney so if you dont deploy something that need the newer requirement you can still use version below g-r. that said our gate jobs always run with whatever is allowed by u-c
19:13:57 dansmith we still have to be below u-c
19:14:11 mriedem u-c bumps are automated
19:14:14 mriedem and merged in batches
19:14:25 openstackgerrit Eric Young proposed openstack/nova master: Support extending attached ScaleIO volumes https://review.openstack.org/554679
19:14:31 dansmith my feeling is that we'll still have some things where you can't actually use nova at the version pinned by cinder for a given thing
19:15:03 mriedem yeah. forget i asked. i need to find a hole to put my head.
19:15:57 mriedem the foot of snow we got yesterday doesn't help
19:16:15 dansmith better you than me
19:19:17 kashyap [nag] Since I'm awake at this random hour in front of the screen, might as well point out this if you are looking for something to +A: https://review.openstack.org/#/c/534384/
19:19:27 kashyap It already has a +2
19:20:29 smcginnis mriedem: I don't think g-r matters anymore and is probably going away at some point.
19:20:40 smcginnis Things just need to be between lower-constraint and upper-constraint.
19:21:52 sean-k-mooney smcginnis: isnt g-r the lower-constraint
19:22:08 sean-k-mooney smcginnis: or at least openstack wide lower-constraint
19:22:11 smcginnis sean-k-mooney: No, there is a new lower-constraint.txt file.
19:22:25 smcginnis sean-k-mooney: Well, I guess you could call that the new g-r.
19:23:01 smcginnis It's slightly different though: https://github.com/openstack/requirements/blob/master/lower-constraints.txt
19:23:13 sean-k-mooney smcginnis: oh ok is that updated automatically some how? just not sure what the delta is between it and g-r
19:23:36 sean-k-mooney oh its expcitly === with no ranges
19:24:12 smcginnis Right. It's saying "the minimum required is exactly this" rather than "it needs to be above this version, but not this one, etc."
19:24:15 sean-k-mooney so in theroy we could use it to test with minium supported version
19:24:27 smcginnis sean-k-mooney: Yep, I think that's the plan.
19:25:21 smcginnis It's a bit of a long read, but Doug wrote out the whole plan here: http://lists.openstack.org/pipermail/openstack-dev/2018-March/128352.html
19:26:29 sean-k-mooney smcginnis: oh good to know. i basically assume that since we only tested with what upperconstraties allowed that anything lower might work but not worth the heart ache of finding out
19:27:09 smcginnis sean-k-mooney: Hah, yeah. And I think in a lot of cases, our lower bound didn't/doesn't accurately reflect what really is the minimum required.
19:27:18 smcginnis This makes it plausible to have a test that can verify that.
19:28:16 sean-k-mooney smcginnis: well the distros. esspcially centos et al would be happy with knowing that a minium is actully checked before they start intergrating
19:28:28 sean-k-mooney ill give the ml post a read thanks
19:28:54 smcginnis sean-k-mooney: Pour yourselve a nice cup of tea first - it will take a while. ;)
19:52:45 openstackgerrit Lee Yarwood proposed openstack/nova master: libvirt: Block swapping to an encrypted volume when using QEMU to decrypt https://review.openstack.org/544238
19:59:44 openstackgerrit Chris Dent proposed openstack/nova master: Move test_report_client out of placement namespace https://review.openstack.org/558911
20:20:03 openstackgerrit Chris Dent proposed openstack/nova master: [placement] Fix incorrect exception import https://review.openstack.org/558916
20:20:35 cdent efried, mriedem, dansmith : that's ^ a fun little bug fix that would be nice to have
20:24:06 openstackgerrit Surya Seetharaman proposed openstack/nova master: Add --enable and --disable options to nova-manage update_cell https://review.openstack.org/555416
20:24:57 mriedem cdent: we can't even have a simple unit test for that if gabbi won't cover it?
20:25:35 cdent mriedem: we can have a unit test for it but it would be...very mocky
20:25:49 mriedem that's fine, it's testing error handling
20:25:52 melwitt yeah, I was about to ask that. I'm not seeing unit tests in the tree for the aggregate handler, guess there hasn't been need of one yet
20:26:12 cdent melwitt, mriedem : as a general rule we haven't done unit tests for the handler code
20:26:16 melwitt that is, gabbit tests cover most of it
20:26:20 melwitt gabbi
20:26:22 mriedem cdent: i know
20:26:25 mriedem but...
20:26:30 mriedem clearly there is a need for it in some cases
20:26:36 melwitt aye
20:26:49 cdent I'm not saying I'm agin it, just that it hasn't happened yet
20:27:02 mriedem you can be a trailblazer here
20:27:19 cdent I was hoping to go to sleep instead
20:27:30 mriedem blaze that treasure trail in the morning
20:27:44 mriedem or lose sleep over it tonight :)
20:28:04 cdent I will lose sleep over trying to generate caring
20:29:00 cdent mriedem: so I can both think about it and not think about it, what is that you're hoping for here?
20:29:43 mriedem i'm -1 without a unit test
20:29:54 mriedem which can be dealt with whenever you feel like it i guess
20:30:00 cdent a test that confirms that the handler raises a 409 when it seens a ConcurrentUpdate, or that a ConcurrentUpdate happens when there is an increment generation failure that casues a ConcurrentUpdate
20:30:18 mriedem i'm fine with the former
20:30:20 mriedem seems easy enough
20:30:25 cdent one of the reasons we haven't done it in the past is because the answer to that ^ is unclear
20:30:35 melwitt +1 on the former
20:30:49 mriedem making sure we don't spew 500 out of the REST API is a simple enough thing to say is a good test
20:31:03 cdent hmmm. Would that even have caught this particular problem?
20:31:35 mriedem the exception moved,
20:31:37 mriedem so yes it should
20:31:37 cdent I suppose so, as the unit test itself would have had an import error when it tried to side effet
20:31:42 melwitt is ConcurrentUpdate a base nova exception too?
20:31:47 cdent not any more
20:31:47 mriedem no

Earlier   Later