Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2017-03-24
16:48:41 johnthetubaguy sdague: thanks I will roll that into the patch I have cooking at the moment
17:02:35 tojuvone johnthetubaguy, Hi I had other things going like floorball. Looks that somehow I did not receive your comments in the patch.
17:04:27 johnthetubaguy tojuvone: added a follow on patch, but I don't really like my ideas in there now, after catch up with sdague on the direction he was suggesting before
17:05:31 tojuvone johnthetubaguy, ok, thanks
17:06:58 openstackgerrit Anthony PERARD proposed openstack/nova master: Add exclusion list for tempest for a libvirt+xen job https://review.openstack.org/449687
17:22:22 openstackgerrit Anusha Unnam proposed openstack/nova master: Add description to policies in simple_tenant_usage.py https://review.openstack.org/449742
17:27:57 EmilienM can we nova ephemeral disk using rbd backend be encrypted with key manager like Barbican? I know it's doable with LVM but not sure about rbd
17:29:22 cdent BobBall_: is this still a thing: https://review.openstack.org/#/c/394857/
17:39:36 tojuvone johnthetubaguy, sdague, bauwser Thanks and sorry if wasted your time around the host maintenance.
17:41:15 cfriesen when running "openstack server migrate" will that do a cold migration? If so, is there any way to do a live migration without specifying the destination compute node?
17:41:26 leakypipes cdent, figleaf, alex_xu, sean-k-mooney: OK, try this on for size: https://review.openstack.org/448282 and https://review.openstack.org/448283
17:41:35 sdague tojuvone: I mostly don't want to waste your time iterating here if we're not going to do it
17:41:51 sdague and I feel like we said we weren't going to do this a while ago
17:43:12 tojuvone sdague, Yes, agree
17:44:39 johnthetubaguy cfriesen: do you try passing host=None or null? it might be an openstack client limitation (https://developer.openstack.org/api-ref/compute/?expanded=live-migrate-server-os-migratelive-action-detail#live-migrate-server-os-migratelive-action)
17:46:29 johnthetubaguy tojuvone: sdague: I agree it probably makes sense for OpenStack to support that use case, its just the preferred implementation/architecture has 99% or 100% of the code changes outside of Nova.
17:46:35 tojuvone sdague, good you stop it there again and can focus doing just what talked already back in Austin without wasting anybody time on something more
17:47:07 johnthetubaguy what we said in austin was?
17:48:07 tojuvone Well, if I recall was around to have something as simple as linking
17:48:48 openstackgerrit Anusha Unnam proposed openstack/nova master: Add description to nova/policies/shelve.py https://review.openstack.org/449763
17:49:14 sean-k-mooney leakypipes: that looks really good to me.
17:49:23 tojuvone Everytim I just touched the review it went bigger and bigger
17:49:43 leakypipes sean-k-mooney: that's more in-line to what you were envisioning?
17:49:56 sean-k-mooney leakypipes: ya it is
17:50:41 johnthetubaguy tojuvone: right, I am not sure the link is useful enough by its-self, disabled host probably is enough for the state in Nova.
17:51:34 openstackgerrit Andrey Volkov proposed openstack/nova master: [placement] Split api-ref topics per file https://review.openstack.org/449764
17:51:35 openstackgerrit Andrey Volkov proposed openstack/nova master: [WIP] Add inventories https://review.openstack.org/449765
17:51:46 sean-k-mooney leakypipes: i still need to fully read trough the register and symbolize function implementations but while its a little more complicated then just a single files with hand coded constances i think is a nicer approch
17:52:26 sean-k-mooney leakypipes: hum they are even simpler then i taught
17:52:42 leakypipes sean-k-mooney: yeah, definitely not rocket science :)
17:52:44 tojuvone johnthetubaguy, if link is set, you know you should read it, if not system is up-to-date
17:53:22 cfriesen johnthetubaguy: looks like openstackclient is fubar'd. there's no way to do a live migration without specifying the dest, and the help text says the default is shared but it's actually defaulting to block.
17:53:30 cdent leakypipes: I've left comments on both. acceptable as is, but still a bit noisy ;)
17:53:31 tojuvone but yes if it is always set, it doesn't help much, but know where to look more detailed info
17:53:40 leakypipes cdent: responded.
17:53:41 figleaf leakypipes: heading out soon, so only a quick perusal so far. I think this is a step closer to what I envisioned, but it still requires more hardcoding than I imagined
17:54:08 figleaf IOW, auto-discovery of the modules and packages, vs. spelling them out
17:54:55 leakypipes figleaf: k. well, I'm happy to check out alternative approaches if you want to hack together something this weekend?
17:54:56 cdent leakypipes: I think a list (without the register) woud make things a lot easier to read: the focus would be on the info (and the associated comments) rather than the code
17:55:24 leakypipes cdent: meh, maybe? :)
17:55:39 cdent it would be more pythonic :)
17:56:41 figleaf leakypipes: ok, don't know if I'll have time, but by Monday fo sho
17:58:26 sdague tojuvone: so... I'm not sure if this helps you at all. But I did a quick exercise this morning of how I was imaging doing these flows outside of nova - https://etherpad.openstack.org/p/host-maintenance-not-in-nova
17:59:21 sdague it really feels like there is a need for a maintenance/status system which users can poll / subscribe to, but I think it lives on it's own
18:01:43 sean-k-mooney leakypipes: im just heading home but ill take a look at it again and play with it over the weekend.
18:02:01 leakypipes sean-k-mooney: cool, thanks!
18:02:03 johnthetubaguy cfriesen: we need to get osc onto the latest microversion, where we default to auto detecting if we need to do block migration or not
18:02:09 johnthetubaguy cfriesen: among other things!
18:03:47 sean-k-mooney leakypipes: cdent on the array vs register i really dont know which i prefer honestly, the register feels a bit like macrocs in c so it make complete sense to me but it may not be the most pythonic approch not sure if an array would be better though. as register could be extended to allow inline documentation
18:04:48 cfriesen johnthetubaguy: yeah, for sure.
18:04:58 sean-k-mooney i think i would be fine with either though as long as we dont have a single file that defines everything
18:05:19 cdent sean-k-mooney: yeah, I think if you're C-oriented it makes good sense, but from a python standpoint it looks like a whole lot of boilerplate. A list (with comments on each item as now) would read well.
18:05:40 cdent mostly I don't want to type register when I come along to add some traits ;)
18:05:53 cdent but yeah, it's reasonably fine as is
18:10:09 tojuvone sdague, thanks, I will take a look. Just home alone with 3 kids, Friday evening at 8 :)
18:10:22 sdague yep, enjoy :)
18:15:20 cdent see everyone in a week, gonna take a few days
18:20:41 superdan mriedem: do you want to recheck this a few times or slam it in? https://review.openstack.org/#/c/449640/
18:23:12 mriedem looking
18:23:47 mriedem let's do it...meow
18:24:04 superdan heh, okay
18:24:40 mriedem got my landscaping buddy on the way over to triage the mini washout
18:24:55 mriedem stopped by ace, the helpful hardware place, on the way home from lunch to stop up on gutter extenders
18:24:57 mriedem booya!
18:25:06 mriedem *stock up
18:25:12 superdan haha
18:27:14 mriedem just trying to keep this dump afloat
18:29:30 sdague are you collecting animals 2 by 2 yet?
18:29:37 mriedem close
18:29:51 mriedem i'll send you a picture
18:30:14 melwitt ace is the place with the helpful hardware folks
18:30:32 mriedem right
18:30:34 mriedem exactly
18:30:44 mriedem the guy that helped me, his name tag said "Lon" and he was missing some teeth
18:30:51 mriedem but god damn did he know a thing or two about downspouts
18:31:26 melwitt perfect
18:31:32 mriedem and the legality of pumping water into the city line
18:31:39 mriedem wink wink thanks lon
18:33:15 sdague heh
18:37:44 openstackgerrit Anusha Unnam proposed openstack/nova master: Add keypairs field to InstancePayload https://review.openstack.org/419730
18:37:54 openstack Launchpad bug 1675791 in OpenStack Compute (nova) "Instance created by demo user(non-admin), shelved by admin and unshelved by demo user --> ends up in error state " [Undecided,Invalid]
18:37:54 mriedem nkorabli: in https://bugs.launchpad.net/nova/+bug/1675791 the error in the conductor log shows that the image used to unshelve the instance can't be found
18:38:01 mriedem so was the image deleted?
18:38:41 nkorabli mriedem: it's just got an error state
18:38:43 mriedem nkorabli: you're hitting htis https://github.com/openstack/nova/blob/master/nova/conductor/manager.py#L665
18:39:16 nkorabli so is it a bug?
18:39:20 mriedem no
18:39:28 mriedem nkorabli: not if you deleted the snapshot image
18:39:53 mriedem when you shelve an instance, the compute API creates a snapshot of the instance: https://github.com/openstack/nova/blob/master/nova/compute/api.py#L3355
18:40:14 nkorabli look into the second screenshot I attached, you can see all the command I run and status of the instance
18:40:17 mriedem that gets stored in the instance metadata when shelving on the compute https://github.com/openstack/nova/blob/master/nova/compute/manager.py#L4326
18:40:39 nkorabli I don't think I deleted any snapshot
18:40:55 mriedem when you unshelve, conductor looks up the snapshot image from that metadata https://github.com/openstack/nova/blob/master/nova/conductor/manager.py#L656
18:41:03 mriedem and would use that image to recreate the instance on another compute
18:41:15 mriedem you're hitting a problem where the snapshot image ID can't be found
18:41:18 mriedem for the shelved instance
18:41:30 openstackgerrit Anusha Unnam proposed openstack/nova master: Add keypairs field to InstancePayload https://review.openstack.org/419730
18:41:55 mriedem nkorabli: do you have access to the database?
18:42:15 mriedem i would check the glance database (or api i guess) for the snapshot image with that id in the error
18:42:22 mriedem the 52d95c... id

Earlier   Later