Earlier  
Later
New Search
Posted Nick Remark
#openstack-nova - 2019-06-06
14:49:48 mriedem "> Yes, it should do exactly that IMHO !"
14:49:52 stephenfin mriedem: How'd you manage to link to a specific email on the list page? Manually adding the anchor (or whatever the #foo part of a URL is called)?
14:50:23 kashyap I totally missed that thread :-(
14:50:30 kashyap Because of e-mail filtering snafu
14:50:33 mriedem stephenfin: "thread" under " Messages sorted by: [ date ] [ thread ] [ subject ] [ author ] ""
14:51:16 stephenfin ooh, nifty
14:51:20 stephenfin mriedem++
14:52:31 sean-k-mooney mriedem: was a blueprint ever filed for that
14:52:38 kashyap Yeah, the thread" view is easy to miss if you're not often parsing the archives :-)
14:53:17 kashyap mriedem: Some elephant-like memory you've got there!
14:53:35 kashyap What is it that you munch on for breakfast?
14:54:27 kashyap sean-k-mooney: No, it wasn't filed, near as I see.
14:56:17 sean-k-mooney thats a shame it would be nice to be able to handel cross arch emultaiton properly in openstack
14:57:31 mriedem sean-k-mooney: not that i'm aware of
14:57:32 kashyap sean-k-mooney: Filing it right now...
14:58:13 mriedem heh so you guys went from "this is a bug burn it burn it!" to "hey it's a feature let's support it!"
14:58:51 mriedem kashyap: i remember the thread because i was the only one engaging chris on it
14:59:04 mriedem and it took me awhile to understand what he was trying to do
14:59:17 kashyap mriedem: I completely missed it due to filtering :-( Normally anything with 'qemu' or 'libvirt' in the thread, I make it a point to engage
15:00:05 kashyap The case is valid for *test* / *devel*: because as of a couple of hours ago, I was running an AArch64 guest on x86_64 -- to test some PCIe stuff
15:00:26 openstackgerrit Lee Yarwood proposed openstack/nova master: WIP libvirt: Use SATA bus for cdrom devices when using Q35 machine type https://review.opendev.org/663011
15:00:26 openstackgerrit Lee Yarwood proposed openstack/nova master: DNM: Run tempest-full-py3 with q35 machine type https://review.opendev.org/662887
15:00:27 lyarwood stephenfin: ^ can you take a look at that during the team call and I'll sort tests out in the background.
15:00:33 kashyap But even for that test to be reliable, I had to ask a person with actual AArch64 hardware
15:00:41 mriedem couldn't the same be said for nested virt?
15:01:19 sean-k-mooney kashyap: test and dev are two of the larger useces for openstack
15:01:45 sean-k-mooney not everything is a long lived NFV app :)
15:01:53 kashyap mriedem: Somewhat; some people use nested for real workloads
15:02:03 mriedem that's my point
15:02:06 kashyap As that still is using hardware extensions
15:02:11 kashyap s/that/that's/
15:02:25 kashyap For pure emulation (or "TCG") -- no, every instruction is emulated
15:02:26 sean-k-mooney some people use emulation for real workload too
15:22:56 stephenfin lyarwood: Done. Think there's _another_ bug here. Might be helpful to get aspiers input on it
15:23:22 stephenfin aspiers: Referring to https://review.opendev.org/#/c/663011/9/nova/virt/libvirt/utils.py@563
15:27:26 lyarwood stephenfin: yup that's true
15:27:55 lyarwood stephenfin: and when you say dedent?
15:28:09 stephenfin the opposite of indent?
15:28:11 kashyap I think he means to unindent
15:28:22 stephenfin oh, I thought dedent was a word
15:28:22 stephenfin TIL
15:28:23 kashyap I got used to "stephenfin speak" on that :D
15:28:35 lyarwood stephenfin: right it is, but I didn't think that was valid tbh
15:28:52 stephenfin the word or what I'm suggesting?
15:29:20 lyarwood stephenfin: what you're suggesting
15:29:35 lyarwood stephenfin: if pep8 is happy then I'm happy, I just didn't think it would be ;)
15:29:49 stephenfin Think I suggested two dedents. Which one are you asking about?
15:30:19 lyarwood stephenfin: the comment is the one I'm looking at
15:31:07 stephenfin the docstring for get_disk_bus_for_device_type? It's indented by 8 but it should only be indented by 4
15:31:32 lyarwood kk
15:32:20 kashyap sean-k-mooney: On "emulation for real workloads": it is completely and utterly baloney because, TCG is completely insecure and upstream provides no guarantee whatsover.
15:32:45 kashyap [It is to be limited to test/dev/CI; that's it.]
15:33:19 sean-k-mooney kashyap: ci is a real workload
15:33:26 kashyap On the "security" bit, of course, some people will come back with: "but, we're in a 'trusted network' [whatever that means]
15:33:50 kashyap sean-k-mooney: Right. I didn't define the word "real", though :-)
15:39:39 kashyap sean-k-mooney: Fresh off the oven:
15:39:41 kashyap < abologna> danpb, kashyap: I tried starting an aarch64/virt guest with 32 pcie-root-ports with ,io-reserve=0 (hacked libvirt) and it boots fine
15:43:42 aspiers stephenfin, lyarwood: dedent is definitely a word ;-) https://docs.python.org/3/library/textwrap.html#textwrap.dedent
15:44:09 aspiers even though maybe not in the official dictionaries ...
16:13:38 openstackgerrit John Garbutt proposed openstack/nova master: Admin password check for project or system scope https://review.opendev.org/663715
16:13:38 openstackgerrit John Garbutt proposed openstack/nova master: WIP: Add new style rule for admin_password https://review.opendev.org/663716
16:13:39 openstackgerrit John Garbutt proposed openstack/nova master: WIP: test admin_password with opt out https://review.opendev.org/663717
16:22:33 openstackgerrit Matt Riedemann proposed openstack/nova master: Update quota known issues docs https://review.opendev.org/662570
16:22:34 openstackgerrit Matt Riedemann proposed openstack/nova master: Cleanup quota user docs https://review.opendev.org/662573
16:43:00 mriedem artom: i dumped some thoughts into your bottom change that renames the nova-live-migration job
16:43:08 mriedem kind of scatterbrained though
16:49:44 openstackgerrit sean mooney proposed openstack/nova master: extend libvirt video model support https://review.opendev.org/647733
16:50:01 openstackgerrit John Garbutt proposed openstack/nova master: WIP: stop admin_password information leakage https://review.opendev.org/663721
16:50:07 sean-k-mooney johnthetubaguy: sorry for the delay but i have added the extra testing to ^
16:50:26 artom mriedem, thank you for your dump (should TYFYD become a thing?)
16:50:59 johnthetubaguy sean-k-mooney: I have to run now, tuba rehearsal later, putting it on the pile
16:51:11 sean-k-mooney no worries
16:51:37 sean-k-mooney am i think it more or less ready to go but we can proably drop it out of the runway at this point
16:53:53 artom I thought they were different operations?
16:54:19 sean-k-mooney they are but they share alot of the same code
16:54:33 sean-k-mooney the only real differens is a cold migrate does not change teh flavor
16:54:40 sean-k-mooney everything else is the same
16:55:24 artom Aha...
16:55:41 sean-k-mooney i think there is some special caseing for resize to the same host but again its minimal
16:56:19 artom mriedem, so, my first thought is, if we already have it in tempest-slow, then keep it there
16:56:32 artom So I'd just abandon https://review.opendev.org/#/c/653498/ entirely
16:57:11 artom But since I don't want to block on https://review.opendev.org/#/c/663405/1, I'd put the fix patch below it
16:57:29 artom So we'd get the revert test coverage with the fix
16:58:32 mriedem artom: cold migrate and resize are the same except for the flavor
16:58:41 mriedem for the intent of your bug fix, it doesn't matter which one runs
16:58:59 artom Conversely, skip the revert tests in tempest-slow entirely, and keep my multinode patch. Because of the -slow part
16:59:15 mriedem artom: i was also thinking that if i can get that cold migrate resize revert test unskipped then yeah we don't need to add new tests to the nova-live-migration job
16:59:21 kashyap stephenfin: mriedem: sean-k-mooney: As promised, I bring you https://blueprints.launchpad.net/nova/+spec/pick-guest-arch-based-on-host-arch-in-libvirt-driver
16:59:38 kashyap [Corrections, snide remarks, rotten tomatoes welcome.]
16:59:40 mriedem artom: i'm not sure how you'd depend on https://review.opendev.org/#/c/663405/ if you drop your job rename patch
17:00:43 artom mriedem, I'd "rebase" my fix below it in order to, in the top patch (aka 663405), still have a revert test with my fix
17:01:08 sean-k-mooney kashyap: there is a way to select host that have hardware accleration already
17:02:51 mriedem artom: you can't rebase a nova change "below" https://review.opendev.org/#/c/663405/ since it's in a different repo...
17:03:20 sean-k-mooney kashyap: you can use the vm_mode image property and specify hvm https://github.com/openstack/glance/blob/master/etc/metadefs/compute-hypervisor.json#L30-L41
17:03:23 artom *facepalm*
17:03:53 artom mriedem, make yours depend on mine then, same effect, right?
17:04:48 sean-k-mooney kashyap: that said we never test that in the ci as far as im aware so no idea if it really works
17:05:44 mriedem artom: that's one option yeah. i'm not sure it's the right option...since unskipping that test should be possible without your change - although i guess one could argue that maybe the test isn't safe if you're using ovs hybrid plugging w/o your fix
17:06:21 artom mriedem, unskipping that change *is* possible without my fix (upstream gate doesn't use hybrid plug)
17:06:43 artom My idea was more to have integration test coverage of my code, at least in the non-hybrid-plug case

Earlier   Later