Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-04-17
15:06:05 mriedem going back to the claims in scheduler vs conductor discussion, it's hard to make that decision without some actual data from both scenarios with a large number of computes
15:06:18 jaypipes mriedem: though it's much much smaller than it used to be...
15:06:44 mriedem i wonder if we could fake things out with a functional test that mocks out 300 computes or something
15:07:05 openstackgerrit Eric Fried proposed openstack/nova master: PowerVM Driver: console https://review.openstack.org/409402
15:07:20 mriedem and then run both options against that - you'd have to do something to introduce random collisions though to force the retry
15:07:39 mriedem didn't yingxin from intel mock something up like that
15:08:17 jaypipes mriedem: kinda I think?
15:11:24 mriedem jaypipes: was thinking of this https://docs.google.com/presentation/d/1UG1HkEWyxPVMXseLwJ44ZDm-ek_MPc4M65H8EiwZnWs/edit?ts=571fcdd5#slide=id.g12d2cf15cd_2_90
15:11:27 mriedem from the austin summit
15:11:48 mriedem alex_xu: is yingxin still on your team?
15:12:24 alex_xu mriedem: yingxin is in the team based on shanghai, he is moving to ceph stuff
15:12:51 alex_xu mriedem: but I remember yingxin put his tool in the github
15:14:31 mriedem https://docs.google.com/presentation/d/1UG1HkEWyxPVMXseLwJ44ZDm-ek_MPc4M65H8EiwZnWs/edit?ts=571fcdd5#slide=id.g12d2cf15cd_2_143
15:14:36 mriedem 1.079 query/secondIn 1000 node simulation
15:14:52 mriedem 99% of the time spent in the scheduler was get_all_host_state
15:15:19 mriedem so we should have dropped that with using placement now for the simple filters
15:15:38 mriedem but it would be interesting to know how much, i.e. 75% now? 50%?
15:16:26 edleafe mriedem: really hard to extrapolate. It would be so nice to have a large test bed to get real numbers from
15:16:42 mriedem edleafe: yeah, if only there was some kind of openstack innovation center...
15:16:54 edleafe R.I.P.
15:16:58 mriedem :(
15:22:08 dansmith mriedem: fwiw, I have a replacement ready for that revert, but will wait until it clears the gate so I don't rebase and kick it out
15:22:43 mriedem ok
15:22:53 mriedem i'm working on a functional test for a regression with filtering servers by tags
15:23:04 mriedem works fine on py27 but fails on py35 of course
15:34:31 openstackgerrit Matt Riedemann proposed openstack/nova master: Add regression test for server filtering by tags bug 1682693 https://review.openstack.org/457103
15:34:31 mriedem dansmith: this ^
15:34:35 openstack bug 1682693 in OpenStack Compute (nova) "tags and not-tags cannot work properly" [High,In progress] https://launchpad.net/bugs/1682693 - Assigned to Zhenyu Zheng (zhengzhenyu)
15:35:11 dansmith ack
16:00:14 openstackgerrit Eric Fried proposed openstack/nova master: PowerVM Driver: SSP emphemeral disk support https://review.openstack.org/443189
16:05:02 clarkb mriedem: edleafe can you not mock it up with a bunch of computes that are just pretending to be scheduled to? I think nova even includes a fake virt driver for this :)
16:05:15 mriedem clarkb: yes
16:05:32 edleafe clarkb: yes, that's what it looks like we will have to do
16:05:34 mriedem clarkb: that was my original idea, do it as a nova-functional test or something
16:06:05 edleafe the main chunk will be the db query and return of matching hosts
16:07:00 edleafe so setting up a reasonable scenario with 1000 hosts and various inventory/allocations should be close enough
16:25:49 mriedem does someone want to put the +W on this novaclient 2.42 change? https://review.openstack.org/#/c/429512/ that's blocking some other pike microversions from getting into novaclient - artom, you might want to review that one too
16:27:43 edleafe Is this doc still correct for cells in ocata? https://docs.openstack.org/ocata/config-reference/compute/cells.html
16:28:05 mriedem edleafe: that's all cells v1
16:28:23 edleafe that's what I thought
16:28:25 mriedem i think i opened a bug for that at one point, sec
16:28:44 mriedem oh nope, similar but different page https://bugs.launchpad.net/openstack-manuals/+bug/1673616
16:28:44 openstack Launchpad bug 1673616 in openstack-manuals "Scaling in Operations Guide - cells section needs to be updated" [Medium,Triaged] - Assigned to Zhenyu Zheng (zhengzhenyu)
16:29:12 edleafe is there a doc for the HostMappingNotFound error?
16:29:33 mriedem edleafe: this? https://review.openstack.org/#/c/456923/
16:30:09 openstackgerrit Jay Pipes proposed openstack/nova master: placement: adds REST API for nested providers https://review.openstack.org/384807
16:30:10 openstackgerrit Jay Pipes proposed openstack/nova master: placement: allow filter providers in tree https://review.openstack.org/377215
16:30:11 openstackgerrit Jay Pipes proposed openstack/nova master: pci: add uuid field to PciDevice object https://review.openstack.org/414768
16:30:11 openstackgerrit Jay Pipes proposed openstack/nova master: placement: integrate ProviderTree to report client https://review.openstack.org/415921
16:30:12 openstackgerrit Jay Pipes proposed openstack/nova master: placement: SRIOV PF devices as child providers https://review.openstack.org/414769
16:30:13 openstackgerrit Jay Pipes proposed openstack/nova master: placement: adds ProviderTree for nested resources https://review.openstack.org/415920
16:30:13 openstackgerrit Jay Pipes proposed openstack/nova master: placement: update client to set parent provider https://review.openstack.org/385693
16:30:14 openstackgerrit Jay Pipes proposed openstack/nova master: placement: add nested resource providers https://review.openstack.org/377138
16:30:45 edleafe mriedem: ok, I'll pass that along
16:30:55 mriedem edleafe: to whom? that's in the ML by the way
16:30:57 mriedem from friday
16:31:19 mriedem edleafe: http://lists.openstack.org/pipermail/openstack-dev/2017-April/115486.html
16:33:30 mriedem here is a bug for the cells config docs page https://bugs.launchpad.net/openstack-manuals/+bug/1683416
16:33:30 openstack Launchpad bug 1683416 in openstack-manuals "Cells in Configuration Reference - warning" [Undecided,New]
16:33:35 mriedem we should expand and clarify that warning
16:33:39 edleafe to someone on my team running the ironic 3rd party CI: http://lists.openstack.org/pipermail/openstack-dev/2017-April/115487.html
16:34:37 mriedem yeah looks like the same problem
16:47:14 openstackgerrit Chris Dent proposed openstack/nova master: [WIP] Register osapi_compute when nova-api is wsgi https://review.openstack.org/457283
16:47:28 cdent sdague (and others): ^
16:47:41 cdent has an important flaw (noted in the commit message)
16:54:38 sdague cdent: so, besides test fixups, what are your concerns there?
16:56:21 sdague because that seems logical enough to me, and all the places where we seem to use this really can't be hit if we didn't ever start an api service
17:42:50 melwitt jaypipes: do we know whenabouts placement stuff will be handling BFV instances properly i.e. not counting them against compute node available disk?
17:43:10 jaypipes melwitt: ? it already should...
17:44:31 melwitt jaypipes: hm, okay. sorry I'm not in the know. I had proposed this in Feb to account for it https://review.openstack.org/#/c/428505/ but let me see if I can figure out if/how it's working now
17:45:35 melwitt that was for claims and reporting. and dansmith had put this one up https://review.openstack.org/#/c/428481
17:46:13 melwitt for requesting zero root disk
17:46:21 jaypipes melwitt: I was thinking of this: https://github.com/openstack/nova/blob/master/nova/scheduler/client/report.py#L147-L154
17:46:52 dansmith claims in the scheduler gets us where we want to be I think
17:47:11 dansmith the hangup with what we have now is the old rt claims for bfv things
17:49:33 melwitt okay. there's old RT reporting for instance usage too but maybe the scheduler is no longer looking at that?
17:49:57 melwitt and then the last piece is whether we request zero root disk for BFV instances
17:49:58 dansmith as of fully-upgraded ocata yeah
17:50:13 dansmith we do request zero for bfv
17:50:16 dansmith already
17:50:38 melwitt okay, nice. so both of those got handled in ocata
17:53:43 melwitt thanks. I'm composing a comment for a BZ about this and wanted to get a ballpark of when things with BFV might be all good
17:58:15 jaypipes melwitt: BZ?
17:58:31 melwitt bugzilla, red hat
17:58:47 jaypipes ah
17:59:42 melwitt there are many customers very interested in this :)
18:00:38 openstackgerrit Nikita Gerasimov proposed openstack/nova master: Set default disk bus for libvirt+Xen HVM to "xen" https://review.openstack.org/457320
18:08:35 openstackgerrit Chris Dent proposed openstack/nova master: [WIP] Register osapi_compute when nova-api is wsgi https://review.openstack.org/457283
18:21:58 mriedem i shiny new donkey to whoever can fix "DeprecationWarning: Policy enforcement is depending on the value of is_admin. This key is deprecated. Please update your policy file to use the standard policy values.'"
18:40:38 openstackgerrit Matt Riedemann proposed openstack/nova master: Use deepcopy when process filters in db api https://review.openstack.org/456872
18:40:42 mriedem jaypipes: i rebased Kevin_Zheng's patch on top of my regression test for the same bug, and fixed something in his patch
18:40:44 mriedem ^
18:42:11 jaypipes mriedem: done
18:42:17 mriedem high 5
18:52:14 jaypipes dansmith: hey, is your cells talk before my talks on resource providers? /me considering adding a "hey, you learned in superdan's talk that X, Y, and Z..." to mine. :)
18:52:36 dansmith jaypipes: heh, I dunnoski
18:53:35 jaypipes dansmith: it is indeed :)
18:54:09 jaypipes dansmith: yours is Wednesday evening from 5:20pm - 6pm, right when everyone's nice and perky. ;)
18:54:37 dansmith jaypipes: yeah, I will be in a super good mood
18:54:39 jaypipes dansmith: mine's the next day around that time when everyone has already left the summit.

Earlier   Later