| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2018-05-11 | |||
| 18:03:56 | prometheanfire | ya, it didn't look bad | |
| 18:06:18 | jgwentworth | cdent: https://bugs.launchpad.net/nova/+bug/1770718 | |
| 18:06:20 | openstack | Launchpad bug 1770718 in OpenStack Compute (nova) "test_auth_middleware_factory fails with new keystonemiddleware 5.1.0" [Undecided,New] - Assigned to Chris Dent (cdent) | |
| 18:06:32 | cdent | thanks | |
| 18:06:46 | jgwentworth | thanks for fixing it :) | |
| 18:18:43 | openstackgerrit | Chris Dent proposed openstack/nova master: Flexibly test keystonmiddleware in placement stack https://review.openstack.org/567923 | |
| 18:18:57 | cdent | jgwentworth, prometheanfire ^ that oughta do the trick (and work for old and new) | |
| 18:19:46 | jgwentworth | saweet | |
| 18:20:47 | openstackgerrit | sean mooney proposed openstack/os-vif master: add noop plugin https://review.openstack.org/567924 | |
| 18:20:58 | prometheanfire | nice | |
| 18:32:18 | mriedem | blarg, if the libvirt driver raises RequestedVRamTooHigh during spawn we'll reschedule to at least 2 other nodes and fail on the same thign | |
| 18:32:19 | mriedem | *thing | |
| 18:33:59 | mriedem | also, double blarg, this is something we could validate in the API | |
| 18:34:36 | jgwentworth | yeah, I was going to ask, do we validate the other extra specs in the API or where | |
| 18:36:13 | openstackgerrit | Merged openstack/nova master: libvirt: always pass emulator threads policy https://review.openstack.org/561913 | |
| 18:39:14 | openstackgerrit | Merged openstack/nova master: Add connection_parameters to list of items copied from database https://review.openstack.org/567248 | |
| 18:42:26 | mriedem | jgwentworth: validating this in the api for server create and rebuild would be trivial, | |
| 18:42:51 | mriedem | only question about that is currently only the libvirt driver supports the image property and flavor extra spec, but i guess that just means - you shouldn't be using these if yo'ure not using libvirt, | |
| 18:43:09 | mriedem | and if you're doing mixed compute, make sure you use host aggregates to pin the image/flavor to libvirt hosts | |
| 18:45:03 | jgwentworth | you mean as far as where the checks belong in the api since they're driver-specific? | |
| 18:45:14 | jgwentworth | s/where/whether/ | |
| 18:47:11 | mriedem | kind of yeah | |
| 18:47:20 | artom | Expose it as traits? | |
| 18:47:21 | mriedem | my point being, the admin controls the flavors, | |
| 18:47:37 | openstackgerrit | Matt Riedemann proposed openstack/nova master: Don't reschedule on RequestedVRamTooHigh errors https://review.openstack.org/567929 | |
| 18:47:38 | mriedem | and if the cloud doesn't have libvirt hosts to support this image property, then don't put that extra spec in the flavor | |
| 18:47:48 | mriedem | and if you have mixed compute, you would need to isolate using aggregates | |
| 18:48:16 | mriedem | artom: that's way more work than i want to put into this :) | |
| 18:48:41 | artom | mriedem, that's fair - I was only saying that as a way to procrastinate from doing real work :D | |
| 18:49:24 | mriedem | you thought, "how can i rope placement into this?" | |
| 18:49:40 | jgwentworth | yeah | |
| 18:50:13 | mriedem | next time i have weird muscle pain i'm going to ask how placement can fix it for me | |
| 18:50:31 | jgwentworth | solid plan | |
| 18:51:28 | artom | Isn't that what chiropractors do? Place your joints and muscles and tendons in all the right, err, places? | |
| 18:52:15 | mriedem | for about 3 days? | |
| 18:52:16 | mriedem | that's what i've heard | |
| 18:52:29 | mriedem | them your body has a periodic task that goofs it all up again | |
| 18:52:32 | mriedem | *then | |
| 18:52:42 | mriedem | similar to our resource tracker | |
| 18:52:43 | mriedem | :) | |
| 18:52:50 | prometheanfire | could be, I noticed that posture and back exercises help a bunch | |
| 18:53:01 | mnaser | so um we're going queens for control plane this weekend most likely | |
| 18:53:06 | mnaser | anything fun to watch out for? | |
| 18:53:19 | prometheanfire | mnaser: it's working fine here | |
| 18:54:05 | mnaser | prometheanfire: upgrade from pike? | |
| 18:56:32 | prometheanfire | ya | |
| 18:56:42 | prometheanfire | old grizzly install originally | |
| 18:59:14 | mriedem_afk | vdrok: recreated that live migration failure http://logs.openstack.org/60/567860/4/check/nova-live-migration/8993c7b/logs/subnode-2/screen-n-cpu.txt.gz?level=TRACE | |
| 18:59:38 | artom | mriedem_afk, huh, so https://github.com/openstack/nova/blob/18.0.0.0b1/nova/virt/libvirt/driver.py#L4703-L4708 is pretty straightforward and doens't need to be in libvirt | |
| 19:00:14 | artom | mriedem_afk, I'm wondering if we can't forgo your small fix and jump straight to in-API validation? It's just comparing image props and flavor extra specs, both are accessible to the API... | |
| 19:00:54 | cfriesen | artom: at what point would you check them? boot/resize/rebuild ? | |
| 19:01:01 | artom | cfriesen, yes. | |
| 19:01:03 | artom | >;) | |
| 19:01:41 | vdrok | mriedem_afk: yup, that's the one. lyarwood 's patch seems to solve it | |
| 19:01:44 | cfriesen | I think those are the applicable three since we need to have both flavor/image to get a full picture | |
| 19:01:50 | artom | cfriesen, actually yeah, good point. We'd need to pick up on every time the image/flavor combo changes. | |
| 19:03:18 | artom | Which isn't a massive deal - I think boot+resize+rebuild actually covers it. But yeah, not as trivial as I thought. | |
| 19:15:19 | openstackgerrit | Jay Pipes proposed openstack/nova master: placement: always create consumer records https://review.openstack.org/567678 | |
| 19:15:20 | openstackgerrit | Jay Pipes proposed openstack/nova master: add consumers generation field https://review.openstack.org/557958 | |
| 19:15:21 | openstackgerrit | Jay Pipes proposed openstack/nova master: placement: Allocation.consumer field https://review.openstack.org/565405 | |
| 19:15:22 | openstackgerrit | Jay Pipes proposed openstack/nova master: rework allocation handler _allocations_dict() https://review.openstack.org/565407 | |
| 19:15:23 | openstackgerrit | Jay Pipes proposed openstack/nova master: Add a microversion for consumer generation support https://review.openstack.org/565604 | |
| 19:47:40 | fried_rice | Hey folks, was pip10 blacklisted in queens? | |
| 19:48:01 | fried_rice | never mind | |
| 20:19:03 | jgwentworth | guh, forgot to update my instance group patch | |
| 20:26:05 | mriedem_afk | cfriesen: that's why i left the api validation to a separate change | |
| 20:26:11 | mriedem_afk | but yes it would be create, resize and rebuild i think | |
| 20:32:31 | arvindn05 | mriedem: can you add https://review.openstack.org/#/c/560596/ to your review list? | |
| 20:33:10 | arvindn05 | i got efried to approve the changes...so need 1 more :) | |
| 20:33:46 | mriedem | vdrok: i'll rebase the recreate patch on top of lyarwood's change once the other thing underneath mine is merged | |
| 20:35:53 | mriedem | arvindn05: yes | |
| 20:41:57 | openstackgerrit | sean mooney proposed openstack/os-vif stable/queens: fix tox py27 job https://review.openstack.org/567942 | |
| 20:53:03 | openstackgerrit | Merged openstack/nova master: Fix irrelevant-files in nova-dsvm-multinode-base https://review.openstack.org/567884 | |
| 20:55:24 | mriedem | arvindn05: done | |
| 20:59:10 | openstackgerrit | Matt Riedemann proposed openstack/nova master: DNM: test raw image live migration with vfat config drive https://review.openstack.org/567860 | |
| 21:05:06 | arvindn05 | mriedem: awesome....thank you :) | |
| 21:05:37 | mriedem | now get cracking on that rebuild patch | |
| 21:05:39 | mriedem | and docs | |
| 21:06:00 | arvindn05 | yep...will have the rebuild patch sometime next week | |
| 21:15:38 | openstackgerrit | Matt Riedemann proposed openstack/nova master: WIP: xenapi: drop deprecated vif_driver config option https://review.openstack.org/567877 | |
| 21:32:18 | openstackgerrit | Merged openstack/os-vif master: fix tox py27 job https://review.openstack.org/567895 | |
| 21:50:21 | openstackgerrit | Eric Fried proposed openstack/nova master: Granular requests to get_allocation_candidates https://review.openstack.org/515811 | |
| 22:00:42 | openstackgerrit | Merged openstack/nova master: Address feedback from instance_list smart-cell behavior https://review.openstack.org/567643 | |
| 22:00:53 | openstackgerrit | Merged openstack/nova stable/pike: libvirt: check image type before removing snapshots in _cleanup_resize https://review.openstack.org/567625 | |
| 22:10:46 | eandersson | How does other people handle scheduling when you have large machines with NUMA flavors. We primarily schedule large machines that takes up a single NUMA zone. The prroblem is that if someone schedules a tiny flavor it will quickly fragment the fleet of computes. | |
| 22:11:54 | eandersson | We used the inverse of TypeAffinityFilter to stack large flavors, so that computes that has a large VM taking up a single NUMA zone will only ever have that flavor size. | |
| 22:14:20 | cfriesen | eandersson: if you use a negative multiplier for the CPU/RAM/disk weighers, it will result in packing rather than spreading. that'll tend to pack all the smaller ones together | |
| 22:15:00 | cfriesen | look for RAMWeigher at https://docs.openstack.org/nova/latest/user/filter-scheduler.html | |
| 22:16:16 | openstackgerrit | Eric Fried proposed openstack/nova master: Granular requests to get_allocation_candidates https://review.openstack.org/515811 | |
| 22:18:54 | eandersson | cfriesen, we did that before introducing our own flavor stacking, and it did not work great, but can't remember our weights at the time. | |
| 22:19:12 | eandersson | pretty sure we ran with nova.scheduler.weights.all_weighers, but not sure what mulipliers. | |
| 22:22:52 | eandersson | We have a large number of single NUMA zone VMs (20 vcpu), and a few 2/4 vcpu VMs | |
| 22:31:55 | cfriesen | eandersson: could you put the small ones in a server group with soft affinity? | |
| 22:34:21 | openstackgerrit | Merged openstack/nova master: update scheduler to use image-traits https://review.openstack.org/560596 | |
| 22:35:06 | eandersson | I don't think we want to put that burden on our user, and it's difficult to enforce on our scale. | |
| 22:40:40 | eandersson | I am guessing that maybe others may be solving this by having different host aggregates for this type of deployments, but we don't want to have to divide our fleet. | |
| 22:46:04 | eandersson | I'll probably follow up with an email to the operators mailing list, but it's something we want to make sure we know how to deal with as we upgrade to Queens (and later Rocky) | |
| 23:15:05 | openstackgerrit | Tsuyoshi Nagata proposed openstack/nova master: nova improvement of maximum attach volumes more than 26 vols https://review.openstack.org/567472 | |
| 23:59:41 | openstackgerrit | Eric Fried proposed openstack/nova master: Granular requests to get_allocation_candidates https://review.openstack.org/515811 | |
| #openstack-nova - 2018-05-12 | |||
| 00:00:53 | openstackgerrit | Jay Pipes proposed openstack/nova master: Add a microversion for consumer generation support https://review.openstack.org/565604 | |