| *** tetsuro has joined #openstack-placement | 00:08 | |
| *** dklyle has quit IRC | 00:31 | |
| *** dklyle has joined #openstack-placement | 00:31 | |
| *** takashin has joined #openstack-placement | 00:42 | |
| *** dklyle has quit IRC | 00:46 | |
| *** mriedem_afk has quit IRC | 01:11 | |
| *** rubasov has quit IRC | 02:24 | |
| *** rubasov has joined #openstack-placement | 02:31 | |
| *** alex_xu has quit IRC | 03:32 | |
| *** alex_xu has joined #openstack-placement | 03:36 | |
| openstackgerrit | Merged openstack/placement master: Link to install docs from deployment overview https://review.openstack.org/643123 | 04:06 |
|---|---|---|
| openstackgerrit | Merged openstack/placement master: Indicate existince of `sync_on_startup` option https://review.openstack.org/643124 | 04:06 |
| openstackgerrit | Merged openstack/placement master: Clean up around links to database migrations scripts https://review.openstack.org/643125 | 04:06 |
| openstackgerrit | Merged openstack/placement master: Clean up the intro to the REST API section https://review.openstack.org/643126 | 04:06 |
| *** alex_xu has quit IRC | 06:06 | |
| *** alex_xu has joined #openstack-placement | 06:07 | |
| *** e0ne has joined #openstack-placement | 07:45 | |
| egonzalez | FYI, kolla patches for placement extraction are being merged. hopefully we have everything in tree today | 08:00 |
| *** takashin has left #openstack-placement | 08:01 | |
| *** tssurya has joined #openstack-placement | 08:12 | |
| *** ttsiouts has joined #openstack-placement | 08:37 | |
| *** helenafm has joined #openstack-placement | 08:44 | |
| *** e0ne has quit IRC | 09:43 | |
| *** e0ne has joined #openstack-placement | 09:46 | |
| *** cdent has joined #openstack-placement | 10:11 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: WIP: Optionally run a wsgi profiler when asked https://review.openstack.org/643269 | 10:42 |
| openstackgerrit | Tetsuro Nakamura proposed openstack/placement master: Trivial: Update doc for _set_traits() https://review.openstack.org/643286 | 10:59 |
| *** ttsiouts has quit IRC | 11:44 | |
| cdent | tetsuro: if you're inclined to approve https://review.openstack.org/#/c/641500/ instead of -1 it, I can move those tests in a follow up. No worries if you would prefer it change now, but if you're okay with it, that stack can start falling in | 11:59 |
| cdent | (I left a comment on the review saying the same thing) | 12:00 |
| tetsuro | cdent: +W'ed the patch at the bottom. I'd like to start from https://review.openstack.org/#/c/642111 tomorrow if I have time and it is still open tomorrow. | 12:16 |
| cdent | Great, thanks. I'm starting that fup patch now. | 12:16 |
| cdent | assuming jay doesn't come round, I assume it will still be open | 12:17 |
| *** ttsiouts has joined #openstack-placement | 12:17 | |
| *** helenafm has quit IRC | 12:30 | |
| *** e0ne has quit IRC | 12:31 | |
| openstackgerrit | Chris Dent proposed openstack/placement master: Move set_traits tests back to test_resource_provider https://review.openstack.org/643313 | 12:32 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Remove unused table constants from allocation_candiate.py https://review.openstack.org/643314 | 12:32 |
| tetsuro | Have a fabulous Thursday. | 12:32 |
| *** tetsuro has quit IRC | 12:32 | |
| efried | egonzalez: That's great news, thanks for the heads up | 12:39 |
| efried | cdent: not sure how this works... [eRxT] | 12:39 |
| efried | [eRxT] | 12:39 |
| cdent | efried: [t eRxt] | 12:40 |
| efried | okay. and... | 12:40 |
| cdent | hmmm | 12:40 |
| efried | why no link? | 12:40 |
| efried | http://p.anticdent.org/eRxT | 12:40 |
| efried | also, when I tried to browse the logs, I had to manually twiddle the querystring to find that message. "earlier" and "later" skipped around too far. | 12:41 |
| cdent | [t eRxT] | 12:41 |
| purplerbot | <egonzalez> FYI, kolla patches for placement extraction are being merged. hopefully we have everything in tree today [2019-03-14 08:00:03.410675] [n eRxT] | 12:41 |
| cdent | it's case sensitive | 12:41 |
| efried | did I do the wrong case? | 12:41 |
| cdent | yes, earlier and later are broken: https://github.com/cdent/purpler/issues/11 | 12:41 |
| efried | okay. | 12:41 |
| cdent | I did the wrong case, but you didn't do the initial t | 12:41 |
| cdent | there's also | 12:41 |
| cdent | [l eRxT] | 12:41 |
| purplerbot | http://p.anticdent.org/eRxT | 12:41 |
| efried | all of which to say: kolla patches for ... | 12:42 |
| cdent | I haven't bothered to fix earlier and later because other things seemed to be more important and nobody was clamouring. patches accepted etc, but also: meh | 12:42 |
| cdent | yay! to the kolla changes | 12:43 |
| *** dklyle has joined #openstack-placement | 13:01 | |
| *** mriedem has joined #openstack-placement | 13:08 | |
| *** e0ne has joined #openstack-placement | 13:32 | |
| openstackgerrit | Merged openstack/placement master: Move Trait and TraitList to own module https://review.openstack.org/641500 | 13:45 |
| openstackgerrit | Merged openstack/placement master: Update get trait and traits calls that were writers to reader https://review.openstack.org/641501 | 14:04 |
| *** helenafm has joined #openstack-placement | 14:16 | |
| dansmith | I hesitate to start a holy war about this, but... | 14:46 |
| * cdent girds loins | 14:47 | |
| dansmith | has there been any discussion about compute node capabilities for image type support being exposed in traits? | 14:47 |
| dansmith | meaning, | 14:47 |
| dansmith | certain configurations of libvirt can't support qcow images, and I assume vmware or hyperv can't either | 14:47 |
| dansmith | so having a compute node expose "I can do qcow" and "I can do raw" and "i can do vmdk" so we can filter on them would be nice | 14:48 |
| cdent | yes, that does sound nice and does seem like a trait-like thing | 14:48 |
| dansmith | otherwise you schedule to a compute, it downloads an image, decides it can't work and fails | 14:48 |
| dansmith | okay, seemed reasonable to me, but, you know :) | 14:48 |
| cdent | ikr. | 14:48 |
| cdent | I think where that would run into trouble is if someone started talking about manipulating those traits on the fly in some wacky way | 14:49 |
| dansmith | mriedem: as a virt capabilities kinda guy, have any thoughts on that? | 14:49 |
| dansmith | cdent: sure, but that probably goes for most everything | 14:49 |
| * cdent nods | 14:50 | |
| *** ttsiouts has quit IRC | 14:50 | |
| *** ttsiouts has joined #openstack-placement | 14:51 | |
| *** ttsiouts has quit IRC | 14:51 | |
| *** ttsiouts has joined #openstack-placement | 14:52 | |
| cdent | dansmith: who/what is the authority on whether the virt can do that thing? Would it require yet-more-config, or is it something virt driver can determine? | 14:53 |
| dansmith | cdent: in the libvirt qcow case, it's totes determinable from existing setup, and I assume vmware can say "no" across the board | 14:54 |
| cdent | "determinable from existing setup" makes my heart sing | 14:54 |
| dansmith | yeah, it would be better to do it a different way if not | 14:55 |
| dansmith | so, as an example of how bad this currently is, | 14:55 |
| dansmith | libvirt nodes that use ceph can't use qcow images,a | 14:55 |
| dansmith | and the advice is to configure glance to automatically flatten any uploaded image to raw across the board globally to prevent an image from being available which is not usable | 14:56 |
| dansmith | which is pretty crazy, and means you can't have two types of compute nodes without a lot of extra work, and presumably would also be done by mixed-hypervisor people for similar reasons | 14:56 |
| * cdent nods | 14:57 | |
| dansmith | so really, libvirt just needs to say "yes to qcow" unless it's configured for ceph | 14:57 |
| cdent | that would be much tidier | 14:57 |
| efried | cdent, dansmith: sorry I haven't been following this, but I'm +1 on the idea, and it sounds totally doable with information the virt driver already knows. | 15:06 |
| efried | We did things in powervm to dynamically set the "can imagecache" virt driver capability based on which disk driver was in play. | 15:06 |
| efried | and the virt driver capability dict is one-line-of-code-per-trait away from mapping directly to compute RP traits. | 15:06 |
| dansmith | right | 15:07 |
| dansmith | and we can pre-filter for max optimization since we know the image details early | 15:07 |
| efried | because https://review.openstack.org/#/c/538498/ | 15:07 |
| efried | yup | 15:07 |
| dansmith | rogjt | 15:07 |
| dansmith | *right | 15:07 |
| dansmith | if I can get mriedem here to not shit on it, I shall plan to execute | 15:08 |
| mriedem | what now? sorry was out trying to unclog drains since my sump pump is running every 30 seconds | 15:10 |
| dansmith | snow melting? | 15:11 |
| mriedem | yeah, plus off and on rain for 2+ days | 15:11 |
| mriedem | so you're talking about something like this yeah? https://review.openstack.org/#/c/640271/ | 15:11 |
| mriedem | or a problem like ^ | 15:12 |
| dansmith | well, | 15:12 |
| dansmith | probably for a similar reason yeah | 15:12 |
| mriedem | so you want to just expose a virt capability via traits as something like COMPUTE_SUPPORTS_QCOW right? | 15:13 |
| mriedem | and then a placement request filter in nova to pre-filter based on the image type | 15:14 |
| openstackgerrit | Chris Dent proposed openstack/placement master: Extract user and endpoint creation install doc https://review.openstack.org/643363 | 15:14 |
| dansmith | yes | 15:14 |
| dansmith | exactly that | 15:14 |
| mriedem | that seems totally fine to me, and if the config on the virt changes from cow to raw, then you sighup the service which wipes the provider tree cache and we rebuild the traits on the next upt periodic | 15:15 |
| mriedem | granted ^ would f any existing instances on that node for rebuild... | 15:15 |
| mriedem | but it's probably a corner case anyway | 15:15 |
| dansmith | yeah | 15:16 |
| mriedem | i think that's at least a nova blueprint to wrangle the trait in compute and placement request filter stuff | 15:16 |
| mriedem | if you're going to ask about...process | 15:16 |
| dansmith | I'll write a spec anyway | 15:16 |
| mriedem | ok | 15:16 |
| *** mriedem is now known as mriedem_away | 15:24 | |
| *** gibi has quit IRC | 16:38 | |
| *** gibi has joined #openstack-placement | 16:38 | |
| *** e0ne has quit IRC | 16:38 | |
| *** tssurya has quit IRC | 16:45 | |
| *** mriedem_away is now known as mriedem | 16:54 | |
| *** cdent has quit IRC | 17:32 | |
| openstackgerrit | Kashyap Chamarthy proposed openstack/nova-specs master: WIP: Gracefully handle QEMU machine types for guests https://review.openstack.org/631154 | 17:37 |
| *** cdent has joined #openstack-placement | 17:46 | |
| *** e0ne has joined #openstack-placement | 17:46 | |
| *** ttsiouts has quit IRC | 17:55 | |
| *** ttsiouts has joined #openstack-placement | 17:56 | |
| *** ttsiouts has quit IRC | 18:00 | |
| *** e0ne has quit IRC | 18:33 | |
| *** e0ne has joined #openstack-placement | 18:40 | |
| * cdent is out of mental capacity | 18:48 | |
| cdent | good night | 18:48 |
| *** cdent has quit IRC | 18:48 | |
| openstackgerrit | Merged openstack/placement master: Remove the TraitList class https://review.openstack.org/641663 | 19:08 |
| openstackgerrit | Merged openstack/placement master: Trivial: Update doc for _set_traits() https://review.openstack.org/643286 | 19:08 |
| openstackgerrit | Eric Fried proposed openstack/placement master: Extract user and endpoint creation install doc https://review.openstack.org/643363 | 19:18 |
| openstackgerrit | Merged openstack/placement master: Move allocation candidate classes and methods https://review.openstack.org/641788 | 19:28 |
| openstackgerrit | Merged openstack/placement master: Reorder classes and methods in allocation_candidate https://review.openstack.org/641789 | 19:28 |
| *** takashin has joined #openstack-placement | 20:47 | |
| *** helenafm has quit IRC | 20:53 | |
| mriedem | fyi, i've invalidated this nova bug since it should be reported in storyboard https://bugs.launchpad.net/nova/+bug/1817458 | 21:22 |
| openstack | Launchpad bug 1817458 in OpenStack Compute (nova) "duplicate allocation candidates with granular request" [Undecided,Invalid] | 21:22 |
| *** mriedem is now known as mriedem_away | 21:27 | |
| *** e0ne has quit IRC | 22:28 | |
| *** mriedem_away has quit IRC | 23:23 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!