| gmann | sean-k-mooney: bauzas: backport of HyperV, vmware driver deprecation backport https://review.opendev.org/q/topic:deprecate-virt-drivers+status:open | 03:22 |
|---|---|---|
| opendevreview | Sylvain Bauza proposed openstack/nova master: DNM : Test a new compute RPC version with an error https://review.opendev.org/c/openstack/nova/+/899392 | 07:42 |
| bauzas | gmann: ack, looking | 07:45 |
| sahid | dansmith: o/ https://review.opendev.org/c/openstack/nova/+/896512 have you had a chance to look at my response to your comment? I guess no :-) | 08:17 |
| opendevreview | Jan Horstmann proposed openstack/nova master: Respect dedicated network during block migration with `iso9660` config drives https://review.opendev.org/c/openstack/nova/+/899458 | 09:23 |
| bauzas | gibi: sean-k-mooney: given how big will be the conversation on VGPUs, I think we should start discussing this afternoon about either the image type backend problem or the live migration URIs | 09:37 |
| bauzas | and constraint that first topic to be only for 30 mins | 09:37 |
| bauzas | so it would leave 1h30 for VGPUs | 09:38 |
| bauzas | gibi: sean-k-mooney: which one would you guys prefer to be discussed in 30 mins | 09:38 |
| bauzas | ? | 09:38 |
| gibi | bauzas: good question | 10:11 |
| gibi | pas-ha had some comments to the migration_uri item on the etherpad where I would need more details. So if he is not around then we might not be able to full settle the migration_uri topic | 10:12 |
| gibi | so I suggest to by default go with the image type first to let pas-ha to come back with details | 10:13 |
| opendevreview | Pranali Deore proposed openstack/nova master: DNM: Test glance new location api https://review.opendev.org/c/openstack/nova/+/891207 | 10:44 |
| bauzas | gibi: ack | 12:22 |
| * bauzas disappears for 20 mins | 12:22 | |
| opendevreview | Merged openstack/os-vif master: Update master for stable/2023.2 https://review.opendev.org/c/openstack/os-vif/+/894064 | 12:39 |
| kashyap | bauzas: Was this already discussed yesterday? "mdev live-migration in libvirt"? | 13:19 |
| kashyap | Looks like so; at least there's notes in the Etherpad | 13:19 |
| bauzas | kashyap: https://ptg.opendev.org/ptg.html | 13:20 |
| bauzas | we're discussing image type backends | 13:20 |
| kashyap | Yeah, I'm just looking at the minutes. I think that image backend code is a bit delicate to do any deep surgery. | 13:21 |
| kashyap | s/minutes/Etherpad/ | 13:21 |
| kashyap | But deprecating the use_cow and force_raw is fine if they're causing more trouble than it's worth it. | 13:25 |
| gibi | kashyap: vgpu stuff coming up soonish | 13:26 |
| gibi | in couple of minutes I guess | 13:26 |
| kashyap | gibi: Thanks for the heads-up! I need to step out to the library as there's some construction noise near my apartment. | 13:27 |
| kashyap | (I'm interested in the live-migration related discussion on that topic. Back in afew) | 13:27 |
| kashyap | sean-k-mooney: Heh, on your "90 seconds" remark! | 14:25 |
| whoami-rajat | bauzas, hey, here is the cinder meeting room https://www.openstack.org/ptg/rooms/cactus | 15:02 |
| whoami-rajat | hope the nova team is available for the session | 15:03 |
| bauzas | whoami-rajat: we're late, trying to join in a couple of mins | 15:05 |
| whoami-rajat | bauzas, sure, no problem | 15:06 |
| bauzas | see you folks either in the cinder room or at 4pm UTC back in the nova room | 15:15 |
| pas-ha[m] | bauzas: hi there, just to highlight my nick for you | 15:16 |
| bauzas | pas-ha[m]: sold ! | 15:16 |
| bauzas | pas-ha[m]: and thanks for voluteering btw. ! | 15:16 |
| sean-k-mooney | bauzas: ptg done? | 16:04 |
| bauzas | sean-k-mooney: no, still on cinder room | 16:04 |
| sean-k-mooney | oh cinder room | 16:04 |
| sean-k-mooney | cool ill join there | 16:05 |
| sean-k-mooney | i just rejoined ours | 16:05 |
| noonedeadpunk | Hey folks. Sorry for bothering with basic libvirt questions, but I'm slightly /o\. After upgrading some computes from 20.04 to 22.04 I've spotted they do have `LIBVIRT_PRT` tables in iptables that does `MASQUERADE all -- 192.168.122.0/24 !192.168.122.0/24 ` | 16:14 |
| noonedeadpunk | So if VM is in network `192.168.122.0/24` - it get's SRCNAT-ed basically, and goes to the virtual namespace from mgmt IP of compute... | 16:15 |
| sean-k-mooney | we dont use libvirt filtering in nova | 16:15 |
| sean-k-mooney | nova-networks may have but nova with neutron does not use libvirt for filtering | 16:16 |
| noonedeadpunk | do you know from top of your head how to just disable that? As I don't see mentioning `open` or smth like that here: https://libvirt.org/firewall.html#the-virtual-network-driver | 16:18 |
| noonedeadpunk | LIke forward? | 16:19 |
| opendevreview | Hao Wang proposed openstack/nova stable/zed: Decorate only Flavor.get_* methods that execute queries https://review.opendev.org/c/openstack/nova/+/899119 | 16:21 |
| sean-k-mooney | im not sure exactly but downstream we are just not deploying the libvirt networkign driver at all | 16:21 |
| sean-k-mooney | noonedeadpunk: what backend are you useing | 16:22 |
| sean-k-mooney | on the neutron side | 16:22 |
| noonedeadpunk | ovs | 16:22 |
| sean-k-mooney | ovs with ml2/ovs with ovs firewal driver, iptable firewall driver, noop firewall driver or are you using ovs with ovn | 16:23 |
| bauzas | we are back in the nova room | 16:23 |
| noonedeadpunk | ah, firewall driver, iptables_hybrid | 16:24 |
| noonedeadpunk | ml2/ovs | 16:24 |
| noonedeadpunk | but yeah, that stuff is created by libvirt default network. Which is wiped as soon as it's dropped | 16:27 |
| noonedeadpunk | and I believe it's because of using iptables_hybrid actually... And what catched me is that before libvirt wasn't creating a default net on install | 16:40 |
| sean-k-mooney | ya you cand safely delete the default network | 16:54 |
| sean-k-mooney | well default newtork in libvirt not in neutron :) | 16:55 |
| noonedeadpunk | not safely though, as then all instances loose network... | 16:55 |
| noonedeadpunk | but can... | 16:55 |
| sean-k-mooney | so really it should not be applying any rules ot the nova vms even with hybrid plug | 16:56 |
| sean-k-mooney | the interface are type eterhenet or bridge and not assocated with the defautl network | 16:56 |
| johnthetubaguy | bauzas: sorry, only just got your note on here. Thanks for the good discussion on the PCI topics, that was useful. | 17:26 |
| sean-k-mooney | johnthetubaguy: i wanted to ask you one other thing before you go | 17:27 |
| johnthetubaguy | sure thing | 17:27 |
| sean-k-mooney | johnthetubaguy: we can chat more about pci in the future but do you know if the ironic-shard work will get completed in C | 17:27 |
| sean-k-mooney | basically it would be nice to rever the revert but are you or JayF going to work on completing that this cycle | 17:28 |
| johnthetubaguy | ah, I am hoping we can re-propose that once the SDK is fixed up, and I think JayF has got good ways to test that now | 17:28 |
| sean-k-mooney | cool | 17:28 |
| sean-k-mooney | ya that is what i was hoping you were going to say | 17:28 |
| johnthetubaguy | I should re-propose the spec, etc. | 17:29 |
| sean-k-mooney | sure feel free to ping when its up. i dont think there is anythign contovertiol other then the testing section | 17:29 |
| JayF | sean-k-mooney: johnthetubaguy: The goal is, in my mind: 1) add better sharding tempest testing in ironic-proper 2) add sharding tempest tests at nova/ironic api interface (with 'bad' nodes in a shard not under test) 3) merge! | 17:29 |
| sean-k-mooney | before we merge the code this time i think we are going to want to see this in an ironic devstack jobs | 17:29 |
| JayF | johnthetubaguy: AIUI the SDK work is already done | 17:29 |
| dansmith | sean-k-mooney: ++ | 17:30 |
| JayF | johnthetubaguy: TheJulia: and I also are going to chat about looking over the Ironic driver to resolve some bugs, including the mixed use of ironicclient and openstacksdk | 17:30 |
| JayF | which leads to appearently inconsistent version negotiation | 17:30 |
| johnthetubaguy | sean-k-mooney: ++ | 17:30 |
| JayF | and completely muddied all this troubleshooting in the moment during bobcat | 17:30 |
| sean-k-mooney | JayF: ya one of our unspoken goals is to complete moving form ironic client to sdk | 17:30 |
| sean-k-mooney | JayF: stephenfin i belive had a series to do that | 17:31 |
| sean-k-mooney | which we just need to review and or update | 17:31 |
| johnthetubaguy | JayF: ace | 17:31 |
| johnthetubaguy | sean-k-mooney: ah, I remembered we were half way there, I didn't realise the code had been up for that, nice. | 17:31 |
| JayF | TBH; PTG planning for Ironic+TC, plus I'm speaking at SeaGL a week from tomorrow, I'm just swamped, and detailed planning for this piece got pushed until after that talk was done :) | 17:32 |
| sean-k-mooney | JayF: johnthetubaguy https://review.opendev.org/q/topic:bp%252Fopenstacksdk-in-nova | 17:32 |
| sean-k-mooney | i belvie that fully move the ironic driver to the sdk | 17:32 |
| JayF | ack; nice! | 17:33 |
| sean-k-mooney | johnthetubaguy: JayF: stephenfin likely wont have time to continue workign on that but if either of ye do then im happy to review | 17:33 |
| JayF | does that man have useful patchsets for everything hidden somewhere | 17:34 |
| JayF | holy cow \o/ | 17:34 |
| johnthetubaguy | JayF: lets catch up next week on who does what, and checking on getting the testing in place, etc. | 17:34 |
| sean-k-mooney | JayF: we had a soft goal to start doing the convertion last cycle in general to using sdk in stead of project client and stephenfin volenteer to show how to do it for ironic | 17:34 |
| JayF | johnthetubaguy: explicitly not next week. Week after next. | 17:35 |
| johnthetubaguy | JayF: cool | 17:35 |
| JayF | johnthetubaguy: note previous message about being busy until a week from tomorrow :D | 17:35 |
| johnthetubaguy | sean-k-mooney: dumb question, is it best as a new patch or a revert revert in the commit message? | 17:38 |
| sean-k-mooney | johnthetubaguy: honestly either is fine i would proably revert the revert and or cleanup the commit message | 17:48 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!