| Posted | Nick | Remark | |
|---|---|---|---|
| #openstack-nova - 2017-03-31 | |||
| 09:39:42 | lyarwood | johnthetubaguy: understood, volume_api or just another method in the driver bdm? | |
| 09:40:02 | johnthetubaguy | lyarwood: good point, probably bdm I guess | |
| 09:40:11 | johnthetubaguy | I could go either way | |
| 09:40:23 | jaosorior | owalsh: alright, oh, so from what I can see in that LP. It seems that there is still a need for the URI there | |
| 09:40:28 | lyarwood | kk, I'll take a look at that and the attach stuff today | |
| 09:40:39 | jaosorior | owalsh: so, why is was that URI being deprecated in the first place? | |
| 09:41:13 | owalsh | jaosorior: not sure | |
| 09:41:38 | johnthetubaguy | lyarwood: actually how do BDMs fit in the non-libvirt drivers these days? | |
| 09:41:56 | jaosorior | seems like it was quite useful and would save a lot of configuration hassles, as opposed to needing to split the config to several options | |
| 09:42:03 | johnthetubaguy | owalsh: jaosorior: I am curious, which URI? | |
| 09:42:17 | owalsh | johnthetubaguy: live_migration_uri | |
| 09:42:59 | johnthetubaguy | owalsh: oh, so bits of that URL are best defaulted by the driver, based on what sort of live-migration you are doing, if I remember correctly, pkoniszewski has more context on that | |
| 09:43:22 | owalsh | johnthetubaguy: https://bugs.launchpad.net/nova/+bug/1671288 | |
| 09:43:23 | openstack | Launchpad bug 1671288 in OpenStack Compute (nova) "live_migration_uri to live_migration_scheme SSH settings" [Undecided,Confirmed] | |
| 09:43:26 | lyarwood | johnthetubaguy: they should all use bdms and driver bdms now AFAIK | |
| 09:43:42 | johnthetubaguy | lyarwood: OK, fingers crossed | |
| 09:43:43 | owalsh | johnthetubaguy: i need to set the ssh user and key | |
| 09:44:50 | johnthetubaguy | owalsh: I see "live_migration_scheme = None" in that configuration reference | |
| 09:45:26 | jaosorior | johnthetubaguy: so, the main issue is not the scheme or the IP. But the extra configurations one can give as query parameters in the URI | |
| 09:45:36 | jaosorior | johnthetubaguy: which vary depending on the scheme used. | |
| 09:47:08 | johnthetubaguy | jaosorior: owalsh: so the aim was to make it easier to use, so the defaults are correct for all the different hypervisors, if thats not going to work out, totally up for reversing the deprecation | |
| 09:47:47 | johnthetubaguy | jaosorior: owalsh: could you just write out like a before and after, to see how bad it looks? | |
| 09:48:32 | pkoniszewski | so live_migration_uri is mostly used to hack QEMU rather than for "extra configuration" | |
| 09:49:16 | pkoniszewski | I understand your problem but I'd like to know your use case for this ssh key | |
| 09:49:26 | jaosorior | pkoniszewski: I wouldn't say that specifying the known hosts, key locations or ssh keys are "hacking qemu" | |
| 09:49:34 | pkoniszewski | does QEMU use root's ssh key? not sure i remember that correctly | |
| 09:50:25 | owalsh | pkoniszewski: it's libvirt, running as root ,so yes | |
| 09:50:28 | pkoniszewski | jaosorior: so mostly operators used this URI to force QEMU to use different link for live migrations | |
| 09:50:44 | pkoniszewski | not to specify keys etc. | |
| 09:51:24 | owalsh | pkoniszewski: I'm adding SSH key setup for the nova user for cold-migration, want to resuse this for live migration to DRY | |
| 09:51:45 | pkoniszewski | got it | |
| 09:51:54 | johnthetubaguy | ah, that makes sense | |
| 09:52:13 | johnthetubaguy | sounds like we should do that out of the box | |
| 09:52:21 | owalsh | pkoniszewski: so qemu+ssh://nova@%s/system?keyfile=/var/lib/nova/.ssh/id_rsa | |
| 09:53:57 | johnthetubaguy | pkoniszewski: whats the best way to do that QEMU? | |
| 09:53:58 | pkoniszewski | well, adding conf option to specify username for SSH makes sense to me | |
| 09:54:21 | pkoniszewski | by default it uses root, this might be a security issue for ome | |
| 09:54:38 | johnthetubaguy | feels like we are adding lots of options at that point... but that might be easiest. | |
| 09:55:24 | pkoniszewski | we can do any, teach people (or let them google it) how to configure URI by themselves, or break it into few options | |
| 09:55:24 | johnthetubaguy | owalsh: I guess migration just uses that key because of the user you are running Nova with, its not a config as such? | |
| 09:55:52 | johnthetubaguy | yeah, I am just curious what folks would prefer | |
| 09:56:06 | owalsh | johnthetubaguy: yea, /var/lib/nova is the nova home dir | |
| 09:56:19 | pkoniszewski | by default LM uses TCP, SSH is just an option | |
| 09:56:25 | openstack | Launchpad bug 1677913 in OpenStack Compute (nova) "percent must be escaped in live_migration_uri" [Undecided,New] | |
| 09:56:25 | owalsh | pkoniszewski, johnthetubaguy: there is also this https://bugs.launchpad.net/nova/+bug/1677913 | |
| 09:56:45 | johnthetubaguy | pkoniszewski: ssh just gives you the secure transport here? | |
| 09:56:57 | pkoniszewski | yes | |
| 09:57:11 | owalsh | johnthetubaguy: ssh + nc is used to tunnel the libvirt unix socket | |
| 09:58:16 | johnthetubaguy | owalsh: that bug about the % needing to be escaped really makes we want us to addd extra config options for the ssl config (and we keep the URL for an extra cycle while that is added) | |
| 09:58:31 | johnthetubaguy | owalsh: would that work for you all? | |
| 09:59:03 | openstackgerrit | Roman Podoliaka proposed openstack/osc-placement master: WIP: implement API version 1.0 https://review.openstack.org/451877 | |
| 09:59:04 | openstackgerrit | Roman Podoliaka proposed openstack/osc-placement master: tests: add a hook for functional testing in the gate https://review.openstack.org/452122 | |
| 10:00:02 | owalsh | johnthetubaguy: yea, might end up with a lot of params though | |
| 10:00:46 | johnthetubaguy | owalsh: but most folks don't need to tweak any of them, and at least it much more discoverable what you can do | |
| 10:01:35 | johnthetubaguy | the alternative seems to involve string substitutions, etc, which is just a bit funky. | |
| 10:05:16 | johnthetubaguy | pkoniszewski: I am curious on the URL substitution, is there a reason why its not the destination host that just provides its whole URL? | |
| 10:06:05 | owalsh | johnthetubaguy: you suggesting we could set it to user@host:port? | |
| 10:07:00 | johnthetubaguy | owalsh: you probably could set that as the host... I guess | |
| 10:07:09 | johnthetubaguy | owalsh: not sure I like that though | |
| 10:07:21 | owalsh | or user@host:port/system?key=foo | |
| 10:07:43 | johnthetubaguy | not sure | |
| 10:07:52 | johnthetubaguy | it might work | |
| 10:08:26 | owalsh | don't think the latter would work as nova will append /system | |
| 10:08:36 | johnthetubaguy | yeah, that sounds right | |
| 10:08:44 | openstackgerrit | Danfly proposed openstack/nova master: remove mox from unit/virt/vmwareapi/test_driver_api.py https://review.openstack.org/452128 | |
| 10:09:10 | johnthetubaguy | owalsh: really I am suggesting we add a new configuration option, and you keep using the URI for now, and we extend the deprecation period of the URI while we get more feedback | |
| 10:10:02 | owalsh | johnthetubaguy: ack, jaosorior ^^ | |
| 10:10:23 | jaosorior | sounds good to me | |
| 10:11:20 | johnthetubaguy | owalsh: jaosorior: would you fancy doing that patch? | |
| 10:11:37 | johnthetubaguy | happy to put it near the top of my review pile once its up | |
| 10:13:20 | owalsh | johnthetubaguy: sure | |
| 10:14:21 | johnthetubaguy | cool, thanks for that info, thats a really interesting use case I had not thought about before | |
| 10:17:53 | owalsh | johnthetubaguy: just thinking...probably need an option for user (ssh only), port (all), and extraparams (all) | |
| 10:18:27 | owalsh | johnthetubaguy: but if user@host:port works with for live_migration_inbound_addr we would just need an extraparams optoin | |
| 10:20:36 | jaosorior | owalsh: I don't see why adding that in live_migration_inbound_addr | |
| 10:20:39 | jaosorior | wouldn't work | |
| 10:20:48 | jaosorior | seems like it's just some substitution | |
| 10:21:15 | owalsh | jaosorior: I was just about to peek at the code :-) | |
| 10:38:22 | johnthetubaguy | jaosorior: owalsh: as long as we clearly document what you tweak to change those things in the comments, we could entertain that | |
| 10:38:39 | johnthetubaguy | johnthetubaguy: owalsh: by we, I mean I | |
| 10:39:28 | owalsh | johnthetubaguy: you're royalty, yea? | |
| 10:39:43 | johnthetubaguy | heh | |
| 10:40:16 | johnthetubaguy | one does ones best to talk like ones queen | |
| 10:40:40 | cdent | one does | |
| 10:51:45 | cdent | Is logs.o.o back to working? | |
| 11:00:00 | openstackgerrit | Jianghua Wang proposed openstack/nova-specs master: Support VGPU https://review.openstack.org/450122 | |
| 11:00:25 | openstackgerrit | Sylvain Bauza proposed openstack/nova master: Destroy the ReqSpec object when deleting the instance https://review.openstack.org/391060 | |
| 11:06:56 | lyarwood | johnthetubaguy: mdbooth is +1 on https://review.openstack.org/439520 btw if you're happy to +W | |
| 11:07:33 | johnthetubaguy | lyarwood: sweet | |
| 11:24:10 | frickler | cdent: recovery of old logs is ongoing, new jobs should be fine | |
| 11:24:21 | cdent | thanks frickler | |
| 11:44:53 | oomichi | gmann: -1 on https://review.openstack.org/#/c/451673 we need to check tempest jsonschema as qa team ;) | |
| 11:58:11 | openstackgerrit | Rodolfo Alonso Hernandez proposed openstack/nova master: Read NIC features in libvirt https://review.openstack.org/444415 | |
| 11:58:12 | openstackgerrit | Rodolfo Alonso Hernandez proposed openstack/nova master: Modify PciPassthroughFilter to accept lists https://review.openstack.org/444946 | |
| 12:14:55 | openstackgerrit | Rodolfo Alonso Hernandez proposed openstack/nova master: Read Neutron port 'binding_profile' during boot https://review.openstack.org/449257 | |
| 12:22:20 | openstackgerrit | Lee Yarwood proposed openstack/nova master: WIP compute: Centralise terminate and detach volume api calls in bdm https://review.openstack.org/452181 | |
| 12:30:55 | openstackgerrit | Andrey Volkov proposed openstack/nova master: [placement] Add api-ref for POST RP inventories https://review.openstack.org/451855 | |
| 12:33:06 | ildikov | johnthetubaguy: hi, around? | |
| 12:33:17 | johnthetubaguy | I am | |
| 12:33:34 | ildikov | just wanted to ask about the meeting slot | |