Earlier  
Posted Nick Remark
#openstack-nova - 2021-03-16
13:51:02 hkominos Will do
13:58:56 openstackgerrit Elod Illes proposed openstack/nova stable/pike: [stable-only] gate: Pin CEPH_RELEASE to nautilus in LM hook https://review.opendev.org/c/openstack/nova/+/780852
14:10:14 stephenfin gibi: I'm reworking Sean's "block unsupported ops with vDPA interface" patch to return HTTP 409 (Conflict) instead of HTTP (Forbidden) since that seems more sensible. It's not a permissions or access issue, IMO. Do you agree?
14:31:57 Luzi hey, is someone around who knows the qemu native tls config options? I think i might have found a bug in the guide: https://docs.openstack.org/nova/latest/admin/secure-live-migration-with-qemu-native-tls.html
14:34:53 kashyap Luzi: Hi, I think I wrote that doc :)
14:35:05 kashyap Luzi: What's the bug? Feel free to amend / send a pull-req :)
14:35:20 kashyap s/pull-req/patch/
14:36:20 kashyap At the bottom right, there is also "found an error? report a bug"
14:36:34 Luzi i followed that guide but got to a point when openstack did not use the tls path, but kept using the tcp path
14:36:56 Luzi i onlny found out because i used tcpdump on both ports
14:37:55 kashyap Luzi: Hmm, if you followed that guide to the word, then your setup should definitely be using the QEMU-native TLS
14:38:49 Luzi i was looking through code and found out, that the uri used for live migration is different from the connection uri and uses a config option not mentioned in the guide
14:39:21 kashyap Luzi: Hmm, possible the guide got slightly outdated ... as migration-related code got reworked
14:39:22 Luzi only after i set live_migration_scheme = tls
14:39:32 Luzi the traffic was encrypted
14:39:44 Luzi and using the correct port :)
14:41:06 kashyap Luzi: I see. That's correct _scheme is favoured
14:41:19 kashyap And _uri parameter is deprecated in favor of the above
14:41:41 Luzi but when you don't set it, as i had not done it, the code uses the hardcoded tcp parameter
14:41:44 fungi looks like that config option was added by https://review.openstack.org/410817 which merged in early 2017
14:42:08 kashyap fungi: Heya; yep. That timeframe is right ... /me clicks
14:42:35 kashyap Yep; patch is also correct.
14:42:43 Luzi live_migration_scheme = tls
14:42:45 fungi first appeared in nova 15.0.0, so ocata and later
14:42:48 kashyap Luzi: Also you don't tell what version of OpenStack you were using?
14:43:13 kashyap Yeah, what fungi says.
14:43:19 Luzi https://github.com/openstack/nova/blob/master/nova/virt/libvirt/driver.py#L1224
14:43:23 openstackgerrit Stephen Finucane proposed openstack/nova master: pci: Add vDPA vnic to PCI request mapping and filtering https://review.opendev.org/c/openstack/nova/+/778350
14:43:24 openstackgerrit Stephen Finucane proposed openstack/nova master: tests: Add functional test for vDPA device https://review.opendev.org/c/openstack/nova/+/780112
14:43:24 openstackgerrit Stephen Finucane proposed openstack/nova master: api: Block unsupported actions with vDPA https://review.opendev.org/c/openstack/nova/+/780333
14:43:25 openstackgerrit Stephen Finucane proposed openstack/nova master: Add release note for vDPA https://review.opendev.org/c/openstack/nova/+/780866
14:43:25 openstackgerrit Stephen Finucane proposed openstack/nova master: WIP: tests: Make mdev stubs work like vDPA https://review.opendev.org/c/openstack/nova/+/780234
14:43:31 stephenfin gibi: lyarwood: ^
14:43:36 Luzi i am working on train and looking to upgrade soon
14:44:12 Luzi i posted the line of code which seems to be still on master
14:45:06 fungi same change deprecated live_migration_uri
14:45:36 kashyap (Yep)
14:46:22 Luzi in the guide the only config option which is mentioned is live_migration_with_native_tls = true
14:46:36 Luzi which is definitely not enough to enable native tls
14:46:40 kashyap Luzi: So, indeed - it checks if the config option is set, then uses the _scheme, if not defaults to TLS
14:47:20 fungi so alternatively, the document could warn that setting the _scheme will prevent a natural fallback to tls
14:47:21 Luzi well i thought it worked, until i started tcpdump
14:47:37 Luzi tcp
14:47:51 Luzi the default in code is tcp
14:48:10 kashyap Luzi: Err, defaults to TCP, I mean, I'm sorry.
14:48:22 Luzi the guid should at least say it is needed to set the _scheme to tls
14:48:38 fungi ayup, i agree it defaults to 'tcp' there
14:49:10 kashyap Luzi: Right; I'll check the installer code to double-confirm; and then can update the doc
14:49:20 fungi okay, so the doc just needs updating to mention setting live_migration_scheme="tls"?
14:49:37 Luzi i'm just worried, that this may also concern some users which did not check their deployment with a tcpdump :D
14:49:51 fungi it got mentioned in the release notes (under features) but yeah that's fairly hidden
14:50:17 Luzi it should be enough, to update the guide, i think
14:50:37 fungi and even the release note doesn't come out and say it's needed for turning on tls
14:50:58 Luzi fungi, yeah thats the point
14:51:06 Luzi i needed a whole day to find out :D
14:51:22 kashyap fungi: Yep - on the doc needs updating. Luzi: Good catch!
14:51:46 fungi Luzi: if you're worried that we need to do some outreach, the typical process for that is an openstack security note: https://wiki.openstack.org/wiki/Security/Security_Note_Process
14:52:00 Luzi i need to go off now, thank you for replying so quickly kashyap and fungi :)
14:52:03 kashyap Luzi: So you need _both_:
14:52:06 kashyap - live_migration_scheme
14:52:14 fungi once published, we can send a copy to the openstack-announce and openstack-discuss mailing lists
14:52:16 kashyap - live_migration_with_native_tls
14:52:44 Luzi fungi, i will do that when i'm at home :)
14:52:47 Luzi kashyap, yes
14:54:57 kashyap (Ah, Luzi is gone before I wanted to mention a way to actually verify how to check native TLS is in effect for disks during migration w/ low-level logs.)
15:03:54 lyarwood stephenfin: https://review.opendev.org/c/openstack/nova/+/780333 - quick commit message nit in there but otherwise LGTM
15:07:19 openstackgerrit Stephen Finucane proposed openstack/nova master: tests: Add functional test for vDPA device https://review.opendev.org/c/openstack/nova/+/780112
15:07:19 openstackgerrit Stephen Finucane proposed openstack/nova master: api: Block unsupported actions with vDPA https://review.opendev.org/c/openstack/nova/+/780333
15:07:20 openstackgerrit Stephen Finucane proposed openstack/nova master: WIP: tests: Make mdev stubs work like vDPA https://review.opendev.org/c/openstack/nova/+/780234
15:07:20 openstackgerrit Stephen Finucane proposed openstack/nova master: Add release note for vDPA https://review.opendev.org/c/openstack/nova/+/780866
15:08:10 lyarwood ta
15:17:37 lyarwood stephenfin: https://review.opendev.org/c/openstack/nova/+/780866 - same in the releasenote btw
15:17:49 stephenfin gdi :)
15:17:59 stephenfin will fix
15:18:36 lyarwood np my review ratio thanks you ;)
15:18:56 lyarwood joys of being a stable core
15:21:31 gibi stephenfin: ack, I have no problem with 409, and I will review the series soon
15:52:32 openstackgerrit Kashyap Chamarthy proposed openstack/nova master: libvirt: Deprecate `live_migration_tunnelled` https://review.opendev.org/c/openstack/nova/+/780908
15:54:27 kashyap gibi: stephenfin: Would be nice to get it going in Wallaby --^. Will reduce some future "live migration config fatigue"
15:55:23 gibi kashyap: ack, I will check after the vdpa series
15:55:38 kashyap Sure; no prob. It's just a deprecation meta-work; that's all.
15:55:53 kashyap I'll add a code-reference in the change comment
15:58:41 lyarwood kashyap: random question, do we have coverage of live_migration_with_native_tls in the gate somewhere?
15:59:15 lyarwood kashyap: I can't find anything but that smells like something we should enable in nova-next at least
15:59:29 kashyap lyarwood: Good question, I do not know; and near as I know, no. It requires a full TLS env to be setup -- which is what TripleO does
15:59:44 kashyap See my prerequisite here: https://docs.openstack.org/nova/latest/admin/secure-live-migration-with-qemu-native-tls.html#prerequisites
16:00:33 kashyap lyarwood: I swear there is some automated test of this upstream (perhaps the upstream whitebox from OSP QE folks), can't find a link yet
16:01:14 lyarwood kashyap: yeah it's likely on the TripleO side and sorry I thought the virt tools would automate the CA creation etc, if not ignore me as it's going to be too much work to enable in devstack for nova-next etc
16:01:54 kashyap lyarwood: Yeah, it requires good old PKI setup. It _can_ be automated, using GnuTLS
16:02:10 kashyap lyarwood: If you don't mind holding your nose, something like ... https://kashyapc.fedorapeople.org/Auto-Setup-GnuTLS/Setup-GnuTLS-CA-and-ServerCert.bash
16:02:55 lyarwood that said we do have some novnc tls stuff in devstack so...
16:03:12 kashyap I see
16:03:36 kashyap But pretty sure I recall chatting w/ TripleO folks to have an automated test. Once I find a URL, I'll link it in the change or post here
16:30:13 bauzas gibi: others, folks, i'm taking an urgent PTO tomorrow, my wife is turning 40 as well
16:32:31 gibi bauzas: ack, have a nice celebration!
16:33:57 gibi bauzas: will you push a new rev from the rpc bump patch before you leave?
16:34:36 bauzas gibi: not sure :(
16:34:53 bauzas still fixing to not support 5.0
16:36:48 gibi bauzas: no worries just preparing my review queue
16:39:34 Luzi kashyap, are you around?

Earlier   Later