Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-03-07
21:58:26 mtreinish mriedem: good waffles do stick together :)
22:07:44 openstackgerrit Sujitha proposed openstack/nova master: [WIP] Reduce calls to load_tags() to 0 https://review.openstack.org/435146
22:08:10 openstackgerrit Sujitha proposed openstack/nova master: Reduce calls to load_tags() to 0 https://review.openstack.org/435146
22:12:20 inc0 hey guys, so we've run into serious issue with new libvirt 2.5
22:12:55 inc0 issue being nova, when creating files for vms (like console.log), uses creates directory with permissions 744
22:13:12 inc0 so if you run nova-compute with different user than qemu, what we do in Kolla
22:13:30 inc0 qemu can't access console.log which makes spawning vm impossible
22:14:02 inc0 and with libvirt 2.5 for reasons unknown to me, dynamic permissiosn seems to be failing to chown these files
22:14:55 inc0 anyone seen that happening? can we specify permissions for vm files (i.e. change them to 774)?
22:32:17 openstackgerrit Sarafraj Singh proposed openstack/nova-specs master: Spec to add per instance timeout for LM https://review.openstack.org/442809
22:33:10 mriedem inc0: you'll probably need to talk to kashyap and/or sgordon on that one, maybe mdbooth - some of the red hat people,
22:33:22 mriedem given the time of day kashyap and mdbooth are probably no longer around, not sure about sgordon
22:33:48 mriedem we're nowhere near testing with libvirt 2.5 in the gate
22:34:18 mriedem we test with 1.3.1 which is what's in xenial
22:34:43 inc0 yes...but not in cloud archive
22:34:49 openstackgerrit Sarafraj Singh proposed openstack/nova-specs master: Add live-migration-force-after-timeout spec https://review.openstack.org/438467
22:34:56 inc0 we will probably fall back to 1.3 ourselves
22:35:13 openstackgerrit Sarafraj Singh proposed openstack/nova-specs master: Spec to add per instance timeout for LM https://review.openstack.org/442809
22:35:38 mriedem inc0: we == kolla?
22:35:46 inc0 mriedem: correct
22:36:22 inc0 I'm not sure when canonical bumped libvirt version this high..but here we are
22:36:24 mriedem ok. i'm not sure how you guys pick what your use or support, but definitely using anything beyond what we test with in the integrated gate using xenial opens you up to risk for stuff like this
22:36:45 mriedem inc0: might as well report a bug regardless
22:37:00 inc0 well yeah, but cloud archive is where you're supposed to pick it up if you install openstack from apt
22:37:07 inc0 and there is libvirt out there too
22:37:15 inc0 not sure if it's libvirt issue or nova issue
22:37:21 mriedem kashyap is good at figuring out what changed in upstream libvirt and why they made such changes, in case they are not backward compatible
22:37:32 inc0 seems more like libvirt tbh..dynamic ownership used to work wel
22:37:34 inc0 l
22:37:44 mriedem i'm assuming it was a conscious decision
22:38:03 mriedem dynamic ownership sounds like a security exposure that they'd close, but i don't know much about it
22:38:05 inc0 kashyap lives in which timezone?
22:38:19 mriedem i think he's in bulgaria?
22:38:22 mriedem dansmith: ^?
22:38:27 dansmith netherlands
22:38:32 mriedem or that
22:38:37 mriedem idk, somewhere in europe :)
22:38:40 mriedem it's all the same right?!
22:38:49 inc0 well, Europe, thanks, I'll try to poke him tomorrow morning
22:38:55 mriedem yw
22:38:56 inc0 no, not the same:)
22:38:59 inc0 but close enough:)
22:39:32 mriedem i asked a russian at the ptg if they vacation in crimea, which apparently was a faux pas...so par for the course
22:40:18 inc0 don't worry, it's getting confusing, in Poland for example time difference to UK changes from 1hr to 2hrs every 6 months
22:40:43 mriedem that's the same in the US
22:40:46 mriedem except for arizona
22:40:51 inc0 daylight savings \o/
22:40:59 inc0 saves no daylight since 1800s
22:41:00 mriedem yeah which is here in another week or so
22:41:15 mriedem it saves me from being able to sleep properly
22:42:38 sdague that reminds me, I really need to find another venue for my "when will then be now" talk about time in software :)
22:43:45 openstackgerrit Eric Brown proposed openstack/python-novaclient master: Remove py34 tox env and pypi classifier https://review.openstack.org/442847
22:43:57 mriedem sdague: as in not AZ?
22:44:34 sdague it included all kinds of fun about dates, timezones, dst, the leapsecond
22:45:51 inc0 duh running qemu with nova user fixes it
22:46:00 inc0 but it's no bueno security
22:47:04 openstackgerrit Matt Riedemann proposed openstack/nova-specs master: List instances using Searchlight (spec) https://review.openstack.org/441692
22:47:10 inc0 anyway guys, please check out libvirt 2.5.0, it potentially breaks lots of stuff
22:48:14 inc0 I mean it's not horrible if only we could make vm files being group-writable
22:50:35 inc0 I think when nova creates dirs for instance it uses default umask which would be 722, if we could make it 772, that would fix it
23:11:07 openstackgerrit Matt Riedemann proposed openstack/nova-specs master: Fix paths when moving specs https://review.openstack.org/432470
23:11:07 openstackgerrit Matt Riedemann proposed openstack/nova-specs master: Fix a couple of spec names https://review.openstack.org/432469
23:11:08 openstackgerrit Matt Riedemann proposed openstack/nova-specs master: Move implemented specs for Ocata https://review.openstack.org/432471
23:17:11 dansmith mriedem: did you see my question about the sort order thing?
23:18:06 dansmith mriedem: I think the sort order is just there to ensure a stable marker search within a single query, not for the overall sort of the results
23:20:41 mriedem looking
23:22:29 dansmith ah yeah diana just confirmed
23:23:43 mriedem me too
23:23:47 mriedem the marker is the uuid
23:23:52 mriedem and i sort of remember talking about this in the spec,
23:24:09 mriedem that uuid being a marker is silly because you can create a new instance and it's random uuid could mess up your marker/sort order
23:24:31 dansmith meh, pagination is lossy anyway
23:24:39 dansmith that doesn't bother me
23:24:59 dansmith so I'm good to push right?
23:25:07 mriedem umm
23:25:12 mriedem so isn't she saying we should be sorting?
23:25:17 mriedem at least by project_id?
23:25:25 dansmith no?
23:25:37 mriedem i'm going to read https://specs.openstack.org/openstack/nova-specs/specs/ocata/approved/paginate-simple-tenant-usage.html quick
23:25:47 mriedem https://specs.openstack.org/openstack/nova-specs/specs/ocata/approved/paginate-simple-tenant-usage.html#data-model-impact
23:26:13 dansmith the only reason we'd need to sort the results is if you care about the sort order as a user right?
23:26:47 mriedem then why was it added to https://specs.openstack.org/openstack/nova-specs/specs/ocata/approved/paginate-simple-tenant-usage.html#data-model-impact?
23:27:05 mriedem https://review.openstack.org/#/c/386771/
23:27:18 dansmith what I have doesn't affect the stability of the sequence across cells
23:27:46 dansmith all that matters is that the last uuid of the end of your limit query lands you in the same place when you go back digging through the cells for the next page
23:29:06 dansmith oh you think she means sorted results to the user _was_ a requirement
23:29:25 mriedem yes
23:29:31 dansmith I read that as "fell out" like, it wasn't in the spec and the code was never updated
23:29:39 dansmith but I guess that's not what she means
23:29:41 mriedem there is a bunch of discussion between alex_xu and diana and jaypipes and johnthetubaguy in https://review.openstack.org/#/c/386771/ about sorting
23:30:04 dansmith so this is even larger of a merge sort than list instances right?
23:30:20 dansmith I mean, this is potentially an order of magnitude or more larger than just a list of live instances
23:31:34 mriedem depends on the window i suppose
23:32:16 mriedem i'm wondering if we only need to sort if detailed=True
23:32:29 dansmith okay then I'm going to move this to later in the set so we can progress on some of the other ones while we argue about this forever
23:32:32 mriedem but, it might also be dependent on the limit/marker
23:33:08 mriedem if you get a tenant_id then all it does is sort on the uuids
23:33:14 openstackgerrit Dan Smith proposed openstack/nova master: Enlighten server tags API about cells https://review.openstack.org/440856
23:33:15 openstackgerrit Dan Smith proposed openstack/nova master: Change MQ targeting to honor only what is in the context https://review.openstack.org/438023
23:33:15 openstackgerrit Dan Smith proposed openstack/nova master: Target cell in super conductor operations https://review.openstack.org/438022

Earlier   Later