| *** igordc has joined #kata-dev | 00:22 | |
| *** auk has joined #kata-dev | 00:45 | |
| *** auk has joined #kata-dev | 00:57 | |
| *** eernst has joined #kata-dev | 04:09 | |
| *** eernst has quit IRC | 04:11 | |
| *** sameo has joined #kata-dev | 05:43 | |
| *** igordc has quit IRC | 05:50 | |
| *** auk has quit IRC | 06:18 | |
| *** sameo has quit IRC | 07:45 | |
| *** sgarzare has joined #kata-dev | 08:05 | |
| *** jodh has joined #kata-dev | 08:08 | |
| *** sameo has joined #kata-dev | 08:27 | |
| *** tmhoang has joined #kata-dev | 08:51 | |
| *** gwhaley has joined #kata-dev | 08:54 | |
| *** davidgiluk has joined #kata-dev | 09:03 | |
| *** sgarzare has quit IRC | 10:19 | |
| *** sgarzare has joined #kata-dev | 10:35 | |
| *** sameo has quit IRC | 11:41 | |
| *** gwhaley has quit IRC | 11:58 | |
| *** tmhoang_ has joined #kata-dev | 12:01 | |
| *** tmhoang has quit IRC | 12:03 | |
| *** gwhaley has joined #kata-dev | 13:07 | |
| *** devimc has joined #kata-dev | 13:10 | |
| *** lpetrut has joined #kata-dev | 13:15 | |
| *** sameo has joined #kata-dev | 13:17 | |
| *** sameo has quit IRC | 13:21 | |
| *** sameo has joined #kata-dev | 13:36 | |
| *** sameo has quit IRC | 14:10 | |
| *** sameo has joined #kata-dev | 14:10 | |
| *** sameo has quit IRC | 14:23 | |
| *** sameo has joined #kata-dev | 14:23 | |
| *** lpetrut has quit IRC | 15:47 | |
| kata-irc-bot | <eric.ernst> @graham.whaley around still? | 16:23 |
|---|---|---|
| kata-irc-bot | <eric.ernst> Do you have a branch kicking around with the kernel config fragments? Or is latest in PR? | 16:23 |
| kata-irc-bot | <graham.whaley> I'm here - latest is in the PR I think. It was failing one of the CIs, and I've not gotten back to look at why | 16:25 |
| kata-irc-bot | <graham.whaley> @eric.ernst ^^ | 16:25 |
| kata-irc-bot | <graham.whaley> @gmmaharaj - looks like maybe podman in the minikube node is not really working - it fails on the 'pivot root' error on the first RUN command, even if that is a mkdir. grrr. I'm giving up on that for now. I will put some notes on my open Issue though. | 16:33 |
| kata-irc-bot | <gmmaharaj> @graham.whaley ohh wow. I am not sure what podman assumes about the OS and the kernel that might hit this. | 16:36 |
| kata-irc-bot | <gmmaharaj> Also wonder if we can add bits to the minikube build when we find out what it is. I guess yes | 16:37 |
| kata-irc-bot | <eric.ernst> FYI: https://github.com/kata-containers/documentation/issues/417 | 16:38 |
| kata-irc-bot | <graham.whaley> oops, accidentally jumped channel. just in case anybody else was watching, the relevant Issue is at https://github.com/kata-containers/packaging/issues/413, where I was trying to make a local kata-deploy image to test | 16:38 |
| kata-irc-bot | <eric.ernst> for "newer folks" I'd *really* like some eyes | 16:38 |
| kata-irc-bot | Action: eric.ernst thinks are design docs could use some love, or at least some top level organizing. | 16:38 |
| kata-irc-bot | <eric.ernst> would be happy to see suggested work items added to that issue. | 16:38 |
| kata-irc-bot | Action: graham.whaley thinks he could easily just take 2 weeks of solid docs writing, if he had the time.... | 16:39 |
| kata-irc-bot | <eric.ernst> i know. | 16:39 |
| kata-irc-bot | <eric.ernst> at least we shuold make sure the docs are linked and referenced. | 16:39 |
| kata-irc-bot | <eric.ernst> Its a major asset of the project, but its been a bit neglected imo. | 16:40 |
| *** igordc has joined #kata-dev | 16:51 | |
| kata-irc-bot | <eric.ernst> @archana.m.shinde - ping? | 17:17 |
| kata-irc-bot | <eric.ernst> how's work on the vsock/kernel issue going? | 17:17 |
| kata-irc-bot | <archana.m.shinde> @eric.ernst @sebastien.boeuf and I figured out the issue | 17:17 |
| kata-irc-bot | Action: eric.ernst goes and checks github | 17:18 |
| kata-irc-bot | <sebastien.boeuf> that was fun | 17:18 |
| kata-irc-bot | <sebastien.boeuf> now someone needs to push the patch | 17:19 |
| kata-irc-bot | <archana.m.shinde> @eric.ernst the kernel patch caused AF_VSOCK to be registered earlier, and so the agent was getting confused | 17:19 |
| kata-irc-bot | <eric.ernst> dumb agent. | 17:19 |
| kata-irc-bot | <eric.ernst> this sounds like a stble candidate patch? | 17:19 |
| kata-irc-bot | <sebastien.boeuf> @eric.ernst oh yes it has to be | 17:19 |
| kata-irc-bot | <archana.m.shinde> yes | 17:19 |
| kata-irc-bot | <eric.ernst> can you paste GH links here? | 17:19 |
| kata-irc-bot | <sebastien.boeuf> 1378 | 17:19 |
| kata-irc-bot | <archana.m.shinde> https://github.com/kata-containers/runtime/issues/1378 | 17:20 |
| kata-irc-bot | <eric.ernst> Ya'll have PR for agent? | 17:20 |
| kata-irc-bot | <archana.m.shinde> so the current proposal is to simplify things and have the runtime explicitly pass info if vsock is to be used or not to the agent via a kernel option | 17:20 |
| kata-irc-bot | <sebastien.boeuf> @eric.ernst the proposed solution involves runtime/agent patches | 17:21 |
| kata-irc-bot | <sebastien.boeuf> but this is the cleanest way long term | 17:21 |
| kata-irc-bot | <sebastien.boeuf> and this simplifies agent logic | 17:21 |
| kata-irc-bot | <eric.ernst> ok, read it now. | 17:22 |
| kata-irc-bot | <eric.ernst> sounds fine. | 17:22 |
| kata-irc-bot | <eric.ernst> what happens if runtime sends cmdline option and it isn't supported yet b y agent? | 17:22 |
| kata-irc-bot | <archana.m.shinde> yeah..I'll find some time today to work on the patch | 17:22 |
| kata-irc-bot | <archana.m.shinde> if anyone else wants to..they can as well :slightly_smiling_face: | 17:23 |
| kata-irc-bot | <eric.ernst> let's rename that issue to be clear, then. | 17:23 |
| kata-irc-bot | <eric.ernst> upstrem vsock fixes break kata | 17:23 |
| kata-irc-bot | <eric.ernst> something like this. | 17:23 |
| kata-irc-bot | <archana.m.shinde> so this should land in our next release | 17:23 |
| kata-irc-bot | <eric.ernst> well, in 1.6.1 and 1.5.4 | 17:24 |
| kata-irc-bot | <eric.ernst> and 1.7.0-alpha-0? | 17:24 |
| kata-irc-bot | <archana.m.shinde> ya | 17:24 |
| kata-irc-bot | <eric.ernst> which is wednseday | 17:25 |
| kata-irc-bot | <eric.ernst> just need to make sure a mix of 1.5/1.6/1.7 doesn't break. | 17:25 |
| kata-irc-bot | <eric.ernst> ie, how do we handle if runtime provides it but guest agent isn't updated to use yet. | 17:25 |
| kata-irc-bot | <eric.ernst> and viceversa | 17:25 |
| kata-irc-bot | <archana.m.shinde> in the meantime there is a workaround to explicitly configure to use vsock in the toml file | 17:25 |
| kata-irc-bot | <archana.m.shinde> let see..if the runtime provides it, but it is an old guest running we cant really do much there | 17:27 |
| kata-irc-bot | <archana.m.shinde> we will fallback to old behaviour | 17:27 |
| kata-irc-bot | <eric.ernst> that's okay - but no new breakage. | 17:27 |
| kata-irc-bot | <eric.ernst> this *only* manifests if they use a particular guest kernel, correct? | 17:28 |
| kata-irc-bot | <eric.ernst> (this is *much* better than the prior vsock failure, then) | 17:28 |
| kata-irc-bot | <graham.whaley> aaw, @raravena80 - nudging @jonolson on his gRPC doc PR is just a cheap thrill eh? :slightly_smiling_face: Can't wait until I next get to do it... | 17:28 |
| kata-irc-bot | <archana.m.shinde> if the runtime does not provide the option, but we have a new guest running, then maybe we can use Stefan's approach as a fallback | 17:28 |
| kata-irc-bot | <archana.m.shinde> @sebastien.boeuf ^ | 17:28 |
| kata-irc-bot | <eric.ernst> its a pretty minimal exposure imo; not sure its worth the complexity? | 17:28 |
| kata-irc-bot | <eric.ernst> our 1.5/1.6/1.7-alpha should include the fixes, and only at that point do we have the updated guest kernel likely, right? | 17:29 |
| kata-irc-bot | <archana.m.shinde> yes, except for clear of course | 17:30 |
| kata-irc-bot | <eric.ernst> stable shuld have the kernel updates anyway, too. | 17:31 |
| kata-irc-bot | <eric.ernst> imo | 17:31 |
| kata-irc-bot | <raravena80> heh, I saw the last nudge was a month ago, so... thought I'd give it a shot :) | 17:32 |
| *** sgarzare has quit IRC | 17:36 | |
| kata-irc-bot | <eric.ernst> OK? https://github.com/kata-containers/agent/issues/506 | 17:36 |
| kata-irc-bot | <sebastien.boeuf> @archana.m.shinde @eric.ernst as a quick fix we can replace the implementation of `VsockSupported()` function with the proposal from Stefan to check for a virtio-vsock device | 17:37 |
| kata-irc-bot | <sebastien.boeuf> as this only involves an agent PR | 17:37 |
| kata-irc-bot | <eric.ernst> what's the downside of *just* doing this? | 17:37 |
| kata-irc-bot | <eric.ernst> i like the cleanliness | 17:37 |
| kata-irc-bot | <archana.m.shinde> @sebastien.boeuf yes thats what I was suggesting above | 17:37 |
| kata-irc-bot | <eric.ernst> and not intorducing dependencies... | 17:37 |
| kata-irc-bot | <sebastien.boeuf> what do you mean? clean? the whole fallback from the agent is not clean | 17:38 |
| kata-irc-bot | <sebastien.boeuf> the runtime has the knowledge about the communication type | 17:38 |
| kata-irc-bot | <sebastien.boeuf> it should be the one providing it to the agent | 17:38 |
| kata-irc-bot | <sebastien.boeuf> the agent should not try to be smart | 17:38 |
| kata-irc-bot | <eric.ernst> what happens if we use vsock for opentracing? | 17:38 |
| kata-irc-bot | <sebastien.boeuf> well the runtime is aware of that | 17:39 |
| kata-irc-bot | <eric.ernst> or for anything else down the road aside from just gRPC? | 17:39 |
| kata-irc-bot | <sebastien.boeuf> well that's orthogonal problem IMO | 17:39 |
| kata-irc-bot | <eric.ernst> i just want to make sure that its generic enough | 17:39 |
| kata-irc-bot | <sebastien.boeuf> you can still add some code for those later | 17:39 |
| kata-irc-bot | <eric.ernst> sounds like it may be - just thinking through other scenarios, and whether agent code seeing vsock is tied to grpc or not | 17:40 |
| kata-irc-bot | <sebastien.boeuf> but here the agent needs to know how to communicate | 17:40 |
| kata-irc-bot | <archana.m.shinde> its not tied to grpc | 17:40 |
| kata-irc-bot | <archana.m.shinde> its about vsock being supported or not | 17:40 |
| kata-irc-bot | <archana.m.shinde> a decision that the runtime can and should make | 17:40 |
| kata-irc-bot | <eric.ernst> cool | 17:40 |
| kata-irc-bot | <archana.m.shinde> and not the agent | 17:41 |
| kata-irc-bot | <sebastien.boeuf> yes | 17:41 |
| kata-irc-bot | <eric.ernst> sounds good to mem. | 17:41 |
| kata-irc-bot | <sebastien.boeuf> but I'm okay if someone wants to push a quick fix to the agent | 17:41 |
| kata-irc-bot | <sebastien.boeuf> since the runtime/agent PR will take at least a week to be merged | 17:41 |
| kata-irc-bot | <eric.ernst> I'd rather see 'real' fix for weds. | 17:41 |
| kata-irc-bot | <sebastien.boeuf> weds? wednesday? | 17:43 |
| kata-irc-bot | <archana.m.shinde> i'll take care of it, i'll open a PR and let you guys comment there | 17:43 |
| kata-irc-bot | <sebastien.boeuf> :+1: | 17:43 |
| *** tmhoang_ has quit IRC | 17:52 | |
| *** jodh has quit IRC | 18:02 | |
| kata-irc-bot | <eric.ernst> TOC is longer than some documents -- i may push to split up the architecture.md | 18:06 |
| kata-irc-bot | <eric.ernst> Friday has turned into docs day for me i guess. | 18:07 |
| kata-irc-bot | <sebastien.boeuf> @eric.ernst I'm going through virtio-fs document, what is the decision about going with one or 2 kernels? Do we think the patches are too "intrusive" not to be applied as part of our default kernel? | 18:08 |
| kata-irc-bot | <graham.whaley> don't know if you can trim the toc to a set number of levels, and if that will help reduce it? So, just top first 2 or 3 levels iyswim | 18:09 |
| kata-irc-bot | <eric.ernst> what's your thought on it, @sebastien.boeuf? | 18:09 |
| kata-irc-bot | <sebastien.boeuf> @eric.ernst I think James and Wei gave +1 to stick to one kernel | 18:09 |
| kata-irc-bot | <eric.ernst> I tend to agree. | 18:09 |
| kata-irc-bot | <eric.ernst> but, we should get #s | 18:10 |
| kata-irc-bot | <eric.ernst> i don't expect it to impact, but... would rather know | 18:10 |
| *** gwhaley has quit IRC | 18:10 | |
| kata-irc-bot | <eric.ernst> managing two is terrible. | 18:10 |
| kata-irc-bot | <gmmaharaj> @eric.ernst for the documentation is there a place you would like me to start with or just jump in head first? | 18:10 |
| kata-irc-bot | <eric.ernst> one is def the way to go .. let's get a PR in place and see the #s | 18:10 |
| kata-irc-bot | <eric.ernst> README.md | 18:10 |
| kata-irc-bot | <eric.ernst> and go fromt here. | 18:10 |
| kata-irc-bot | <eric.ernst> I'm mostly concerned about the documentation of design, though | 18:10 |
| kata-irc-bot | <eric.ernst> less aboutthe process based ones.. for now. | 18:11 |
| kata-irc-bot | <graham.whaley> local numbers with the report tool are best way - more detail than just the metrics CI | 18:11 |
| kata-irc-bot | <sebastien.boeuf> @eric.ernst ok sounds good. And yes I would prefer one kernel only myself | 18:11 |
| kata-irc-bot | <eric.ernst> Thanks for jumping back into the proposal document. | 18:11 |
| kata-irc-bot | <eric.ernst> I'm happy to work through it later today if you want help. | 18:12 |
| kata-irc-bot | <eric.ernst> Goal should be to close out much the of the comments and mark as addressed (oncet hey are addressed) | 18:12 |
| kata-irc-bot | <sebastien.boeuf> @eric.ernst sounds good, do you want to sync at 3? | 18:26 |
| kata-irc-bot | <eric.ernst> happy to | 18:26 |
| kata-irc-bot | <sebastien.boeuf> @eric.ernst remember we were wondering which one to pick between 4.20 and 5.0 "stable"? well I think we have our answer from kernel.org. The stable 4.20.17 is EOL | 18:55 |
| kata-irc-bot | <sebastien.boeuf> so 5.0 it is, right? | 18:55 |
| kata-irc-bot | <eric.ernst> what I propose. | 18:56 |
| kata-irc-bot | <eric.ernst> CL is at 5.0.5 FWIW | 18:56 |
| kata-irc-bot | Action: eric.ernst needsto recall the stable window. And waht the next LTS is planned to be . | 18:57 |
| kata-irc-bot | <eric.ernst> Once new LTS is cut (Q4), we can probably re-evaluate need to hob. | 18:58 |
| kata-irc-bot | <eric.ernst> hop* | 18:58 |
| kata-irc-bot | <sebastien.boeuf> Q4? | 19:34 |
| kata-irc-bot | <sebastien.boeuf> I'm not following here, I thought we wanted velocity to keep tracking latest features | 19:35 |
| kata-irc-bot | <eric.ernst> yes, but perhaps we should look @ specific features | 20:15 |
| kata-irc-bot | <archana.m.shinde> @sebastien.boeuf Just a suggestion, would be useful to describe in a line or two what the kernel patches are meant for, what qemu patches are for in the virtio-fs section of your doc | 20:16 |
| kata-irc-bot | <archana.m.shinde> you mentioned about the patches being non-intrusive | 20:16 |
| kata-irc-bot | <archana.m.shinde> would be good to have that | 20:17 |
| kata-irc-bot | <sebastien.boeuf> yeah I can do that | 20:17 |
| *** davidgiluk has quit IRC | 20:21 | |
| *** sameo has quit IRC | 20:43 | |
| kata-irc-bot | <eric.ernst> FYI: https://github.com/orgs/kata-containers/projects/16 | 20:44 |
| *** igordc has quit IRC | 22:09 | |
| *** igordc has joined #kata-dev | 22:13 | |
| *** devimc has quit IRC | 22:46 | |
| *** igordc has quit IRC | 23:30 | |
| *** igordc has joined #kata-dev | 23:33 | |
| *** dklyle has quit IRC | 23:48 | |
| *** david-lyle has joined #kata-dev | 23:48 | |
Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!