Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2019-10-07
19:00:54 sean-k-mooney mriedem: right fair point i was thinking about usage in nova.
19:03:31 dansmith the good news, at least,
19:03:51 dansmith is that this is all just duck-alignment and that the whole thing works and pre-caches the image as expected
19:04:16 sean-k-mooney :)
19:04:51 sean-k-mooney have you implmented the trotteling code or just the basic poc
19:05:38 dansmith I've done the parallelism=N stuff yeah
19:06:05 sean-k-mooney cool
19:06:08 dansmith I'm going to push up in a sec so I can review it myself and look for tests I'm missing (of which there are many)
19:06:58 sean-k-mooney its funny but i often find the change in media between my ide and the gerrit review helps with things like that
19:07:09 dansmith yes, exactly
19:07:18 openstackgerrit Dan Smith proposed openstack/nova master: Add cache_image() support to the compute rpc, api, and manager https://review.opendev.org/687138
19:07:18 openstackgerrit Dan Smith proposed openstack/nova master: Add cache_image() driver method and libvirt implementation https://review.opendev.org/687137
19:07:19 openstackgerrit Dan Smith proposed openstack/nova master: WIP: Add image caching API for aggregates https://review.opendev.org/687140
19:07:19 openstackgerrit Dan Smith proposed openstack/nova master: Add cache_images() to conductor https://review.opendev.org/687139
19:07:29 dansmith gerrit is where I'm used to looking for those things
19:08:12 openstackgerrit Dan Smith proposed openstack/python-novaclient master: WIP: Add aggregate-cache-images command and client routines https://review.opendev.org/687141
19:09:10 sean-k-mooney by the way you are not storing the cached state of the image anywhere correct? that is just known on the compute nodes
19:09:22 dansmith I'm really tempted to add a "check only" flag so you can single-shot the api and just say "check and see which of these aren't done to give at least the tiniest bit of reporting, but I think I need to resist the urge
19:09:29 dansmith correct
19:09:32 dansmith per the spec :)
19:10:14 sean-k-mooney yep im rebasing my schduler prefilter code to master and just wanted to confirm in my head that it should not conflict
19:10:19 dansmith many of the previous attempts stalled out at the persistence layer so I definitely want to punt on that
19:11:17 sean-k-mooney ya the tangent my brain was one was could you use a weighr to prefer the hosts wher eit was cached by reporting the cached images in the hostSate object
19:11:55 sean-k-mooney anyway not important now
19:12:42 dansmith sean-k-mooney: a weigher for exactly that is one of the ill-fated "previous attempt" items at the bottom of the spec
19:13:19 sean-k-mooney ah i see well there is nothing preventing it being added at a future point in your design
19:13:20 dansmith a sunken treasure for you
19:15:26 dansmith I want to roll this out and hear the requests and justifications for more complicated stuff
19:15:31 sean-k-mooney ah https://blueprints.launchpad.net/nova/+spec/node-cached-image-weigher
19:16:07 sean-k-mooney ya makes sense
19:30:26 openstackgerrit sean mooney proposed openstack/nova master: Libvirt: report storage bus traits https://review.opendev.org/666914
19:30:27 openstackgerrit sean mooney proposed openstack/nova master: Add transform_image_metadata request filter https://review.opendev.org/665775
19:30:27 openstackgerrit sean mooney proposed openstack/nova master: libvirt: use domain capabilities to get supported device models https://review.opendev.org/666915
19:31:24 sean-k-mooney efried: do you mind removing your -2 from https://review.opendev.org/#/c/666914/ now that master is open for ussuri
19:32:08 efried sean-k-mooney: will do, but I'm going to put the bp "Direction" back to pending
19:32:30 efried done
19:32:48 sean-k-mooney efried: ok.
19:33:26 sean-k-mooney so does that mean it can still merger or does it have to wait till m2?
19:35:32 openstackgerrit Eric Fried proposed openstack/nova-specs master: Add 'Feature Liaison' spec process https://review.opendev.org/685857
19:36:26 efried sean-k-mooney: You can still merge code. But if when we have this planning discussion (format/timing/etc still tbd) it is decided we're going to cut it, we'll stop merging code for it.
19:36:43 sean-k-mooney ah ok
19:40:10 efried basically you still have to treat it as if it could end up partially merged, just like we normally do. IOW if it's a tightly-coupled series, we'll put -2 on the bottom until it's ready so it's all or nothing; otherwise we can merge it in stages.
19:40:35 efried btw, updated the core liaison (now feature liaison) patch, in case you were in mid-rebase ----^
19:41:56 sean-k-mooney efried: i upload my spec this morning ill rebase it again and update the section then im going to call it a day
19:42:34 sean-k-mooney oh maybe i did not push that
19:43:36 sean-k-mooney ya i still have it locally
19:50:45 sean-k-mooney i decied to try out http://spacemacs.org/ at the weekend. i have been using it all day and so far i like it
19:51:13 sean-k-mooney i disabled the vim keybindings but it works pretty well for me so far
19:51:42 sean-k-mooney not sure it will replace pycharm but it might replace my use of nano
19:52:14 openstackgerrit sean mooney proposed openstack/nova-specs master: Add spec for vm scoped sriov numa affinity https://review.opendev.org/683174
19:52:32 sean-k-mooney anyway o/
20:17:30 efried mriedem: how do we deal with bugs with a proposed fix that's now abandoned? Move back to the latest unassigned state and remove the owner?
20:19:01 dansmith yes
20:20:03 openstackgerrit Dan Smith proposed openstack/nova master: Add cache_image() support to the compute rpc, api, and manager https://review.opendev.org/687138
20:20:04 openstackgerrit Dan Smith proposed openstack/nova master: WIP: Add image caching API for aggregates https://review.opendev.org/687140
20:20:04 openstackgerrit Dan Smith proposed openstack/nova master: Add cache_images() to conductor https://review.opendev.org/687139
20:29:46 mriedem efried: yeah, there are cases where if i'm familiar with the thing i can just close it as invalid (or fixed by something else)
20:29:53 mriedem if it has to go back through triage or whatever though yeah
20:30:10 efried k, thx
20:31:14 mriedem dansmith: gonna throw your stuff in https://etherpad.openstack.org/p/nova-runways-ussuri ?
20:39:19 openstackgerrit Artom Lifshitz proposed openstack/nova master: Helper to start computes with different HostInfos https://review.opendev.org/686832
20:39:20 openstackgerrit Artom Lifshitz proposed openstack/nova master: [WIP] Functional tests for NUMA live migration https://review.opendev.org/672595
20:39:34 artom mriedem, dansmith ^^ if you get a chance to start taking a look
20:39:47 artom I left the main func tests as WIP because I want to add a test for the thing that ya.wang found
20:41:11 openstackgerrit Artom Lifshitz proposed openstack/nova master: libvirt: Ignore volume exceptions during post_live_migration https://review.opendev.org/682622
20:41:11 openstack bug 1843639 in OpenStack Compute (nova) "libvirt: post_live_migration failures to disconnect volumes result in the rollback of live migrations" [Medium,In progress] https://launchpad.net/bugs/1843639 - Assigned to Lee Yarwood (lyarwood)
20:41:11 openstackgerrit Artom Lifshitz proposed openstack/nova master: Add regression test for bug #1843639 https://review.opendev.org/682621
20:41:44 mriedem artom: just a thought, but you don't need to load up every test ever into the same patch for the numa lm functional stuff
20:42:01 mriedem like, happy path + framework stuff and then build on top in separate changes the weird edge cases and failure/rollback scenarios
20:42:09 mriedem that would make it a lot easier to review as well
20:42:14 openstackgerrit Merged openstack/nova master: Error out interrupted builds https://review.opendev.org/666857
20:42:42 artom mriedem, true
20:42:50 artom It'd make the reviewing more breakdown-able, in any case
20:42:52 artom Not sure about easier
20:43:14 mriedem broken down usually means easier
20:43:47 mriedem it would at least be easier for me to get the motivation to get started if it wasn't 1K in a single change...
20:43:55 artom I was thinking in the sense of "reviewing one test method/class per day" can be done in either case
20:43:56 mriedem a single WIP change at that
20:44:08 mriedem i guess it's not 1K
20:44:22 artom If you'd prefer it split into smaller chunks I can do that
20:44:50 artom But you'll just get 5 200LOC patches with a test class each instead of 1 KLOC patch with 5 test classes ;)
20:45:15 mriedem that's essentially what i did with the functional tests for cross-cell resize
20:45:29 mriedem starting here and then going up https://review.opendev.org/#/c/636253/
20:46:20 mriedem i guess my point is seeing WIP + lots of code makes me really not want to start reviewing it :)
20:46:32 mriedem i will find any reason to procrastinate with something else
20:46:50 artom Fair enough - it's not like I didn't procrastinate :)
20:47:06 artom But... even if I -1 random changes of your for tenuous reasons? ;)
20:47:12 artom *of yours
20:47:49 mriedem you mean that will make me want to review your stuff *more*?
20:47:55 artom It doesn't?!
20:48:05 mriedem it might add to the guilt complex
20:48:10 artom I've been doing it wrong all this time! (by which I mean a week)
20:48:47 mriedem efried: i added cross-cell-resize to the ussuri runways queue, are we able to queue stuff up now?
20:50:09 efried mriedem: might as well.
20:50:39 mriedem efried: ok you could probably drop the -2 on https://review.opendev.org/#/c/674243/ as well
20:51:36 efried done
20:52:08 dansmith mriedem: how does runways work with the new process? I guess I had kinda imagined that we weren't going to do that
20:52:36 mriedem goto efried
20:52:52 efried (how) did it work before?
20:53:03 mriedem i wouldn't think the new spec approval process would change anything about runways

Earlier   Later