Wednesday, 2020-04-01

*** devimc has joined #kata-dev00:49
*** devimc has quit IRC03:18
*** dklyle has quit IRC05:00
*** sameo has quit IRC06:31
*** irclogbot_2 has quit IRC06:49
*** sameo has joined #kata-dev06:50
*** irclogbot_0 has joined #kata-dev06:51
*** irclogbot_0 has quit IRC06:53
*** irclogbot_0 has joined #kata-dev06:53
*** jodh has joined #kata-dev06:54
*** sgarzare has joined #kata-dev07:09
kata-irc-bot1<fidencio> That's weird, I'll check if some of our teams have access to azure and try to connect to one of the instance and try to figure out here what may be causing such issues. That's really weird we're facing those issues in the first place.08:02
*** ailan has joined #kata-dev08:03
*** gwhaley has joined #kata-dev08:05
*** davidgiluk has joined #kata-dev08:05
*** noahm has quit IRC09:05
*** tmhoang4 has joined #kata-dev09:49
*** tmhoang has quit IRC09:50
*** tmhoang4 is now known as tmhoang09:50
*** crobinso has joined #kata-dev10:53
*** gwhaley has quit IRC10:59
*** davidgiluk has quit IRC11:29
*** gwhaley has joined #kata-dev12:03
*** davidgiluk has joined #kata-dev12:35
*** devimc has joined #kata-dev13:34
gwhaleyhi @fuentess @gabyct - I have a minor question - on branch (tag) 1.10.2, I have a feeling we wound back the 'yq' version (to 2.3.0), but maybe we forgot to wind back an invocation in the packaging repo: https://github.com/kata-containers/packaging/blob/1.10.2/scripts/lib.sh#L6014:43
gwhaleyIt is trying to use '-X', which I think is for the new 3.2.0 yq. I happened to run into it trying to do an automated QAT kernel/rootfs build from a tag... just happened to choose one that is maybe broken?14:44
kata-irc-bot1<salvador.fuentes> gwhaley, https://github.com/kata-containers/packaging/blob/1.10.2/scripts/lib.sh#L23 It should be getting the latest one, but if you already have one (older) in your system, it wont install the correct version, I think we may need to change that14:49
*** dklyle has joined #kata-dev14:51
gwhaley@futentes - oh, let me look at that. I don't have a previous one installed afaik as this is a clean run in a container - I ended up with v2.3.0... so let me see how that gets there..... I think that is the version in the versions.yaml file you see on that tag..15:05
gwhaley@fuentess - ah, it is the tests/.ci/install_yq.sh that has the 2.3.0 hard wired.... in that branch/tag15:08
kata-irc-bot1<salvador.fuentes> gwhaley yeah, that is weird, we should have same versions accross the repos15:24
gwhaley@fuentess - let me see if I find the energy to do a backport, but I'm not sure how easy we can test it. I guess the problem I ran into is doing both a tests repo install_yq then followed by a packaging install_yq on the same platform (build).... we probably never do that - we either test or we package I suspect..15:27
kata-irc-bot1<salvador.fuentes> gwhaley, yeah, we'll probably wont hit the issue... at least seems that CI jobs are currently happy15:31
*** pcaruana has quit IRC15:32
*** pcaruana has joined #kata-dev15:48
*** devimc has quit IRC16:00
*** devimc has joined #kata-dev16:01
*** sgarzare has quit IRC16:26
*** gwhaley has quit IRC17:03
*** amorenoz has quit IRC17:03
*** amorenoz has joined #kata-dev17:04
*** jodh has quit IRC17:07
*** noahm has joined #kata-dev17:23
*** errordeveloper has joined #kata-dev18:07
errordeveloperso I've been able to build cutom kata image my own way :tada:18:07
errordeveloperbut there are a few issues:18:07
errordeveloperlooks like kata is relient on a partition scheme, with DAX etc, so you cannot just use any image that qemu can boot18:08
errordeveloperbut I've been able to borrow the logic from osbuilder to do that18:09
errordeveloperwhat is more of a challenge for me is kernel depenendencies18:09
errordeveloperI tried one of the ubuntu bionic kernels and ran into an issue with root device not being available (pmem0)18:10
errordeveloperI came accross https://bugs.launchpad.net/ubuntu/+source/linux-kvm/+bug/176185418:11
openstackLaunchpad bug 1761854 in linux-kvm (Ubuntu Bionic) "kata-containers: Cannot open root device "pmem0p1"" [Undecided,Fix released] - Assigned to Kamal Mostafa (kamalmostafa)18:11
errordeveloperand tried latest -kvm kernel, but that didn't work either18:11
errordeveloperI am now looking to try centos kernels, will see what works18:11
errordeveloperbut I'd rather like to be able to take any distro kernel and make kata use it18:12
*** errordeveloper has quit IRC18:34
*** errordeveloper has joined #kata-dev18:41
fidencioerrordeveloper: CentOS as in CentOS 7 or CentOS 8?18:48
kata-irc-bot1<errordeveloper> @fidencio I tried 8 actually18:48
fidencioerrordeveloper: you'll face some issues with CentOS 8, IIRC18:48
fidencioerrordeveloper: for instance, I don't think 9p is available on their kernel (but I may be mistaken here)18:49
errordeveloperI think I treid 7 also, neither booted18:49
fidencioerrordeveloper: and I don't think virtiofs will be available till next release18:49
errordevelopernamely, I have: vmlinuz-3.10.0-1062.18.1.el7.x86_64 and vmlinuz-4.18.0-147.5.1.el8_1.x86_6418:49
fidencioerrordeveloper: would you be willing to give a try to the Fedora kernel?18:49
errordeveloperfidencio: sure, it would make sense18:50
errordeveloperbut my aim is to enable use of any distro kernel, is that going to be possible somehow?18:50
fidencioerrordeveloper: oh, that may be even easier18:50
errordeveloperwell, say a subset, doen't have to be every single one18:50
fidencioerrordeveloper: you just have to be sure you have all the modules you need18:50
fidencioerrordeveloper: for Fedora, what we're doing is using the script I pointed at your direction last time we talked18:51
errordeveloperfidencio: hm.. well, so what about that ubuntu issue?18:51
fidencioerrordeveloper: sincerely, don't know. not a ubuntu user / developer18:51
errordeveloperI tried the 5.3 kernel from bionic, and it couldn't mount pmem018:51
errordeveloperhttps://bugs.launchpad.net/ubuntu/+source/linux-kvm/+bug/176185418:51
openstackLaunchpad bug 1761854 in linux-kvm (Ubuntu Bionic) "kata-containers: Cannot open root device "pmem0p1"" [Undecided,Fix released] - Assigned to Kamal Mostafa (kamalmostafa)18:51
errordeveloperas that bug suggests, I tried -kvm kernel, and that didn't work either18:52
errordeveloperyou said I just need modules, but modules are in rootfs...18:52
errordeveloperhonestly, I think that statement was a bit underestimated :)18:53
errordeveloperI will try fedora tomorrow18:53
errordeveloperwhich fedora release+kernel I should go for?18:53
fidencioerrordeveloper: just spawn a VM with Rawhide18:53
fidencioerrordeveloper: I've been using rawhide for development and rawhide / f32 have all the needed bits for virtiofs18:54
errordeveloperfidencio: ok, I see18:55
fidencioerrordeveloper: spawn a rawhide machine; dnf -y install kata-runtime; check your /var/cache/kata-containers/18:55
errordeveloperI guess my question is - can I make kata run qemu in most basic mode, without anything special, just the bare mimum?18:56
fidencioI'll leave that for others to reply :-018:57
fidencio:-)18:57
errordeveloperfidencio: well, fedora comes with kernel-4.18.0-147.5.1.el8_1.x86_6419:00
errordeveloperwhich is the same as CentOS 819:00
errordeveloperand that didn't work with kata19:00
errordeveloperfidencio: I think you didn't catch what I'm trying to say, I want kata to boot pods with kernels from various distros19:01
errordeveloperI just really would like to know if that's close to possible, or kata is very dependent on its kernel configs19:04
fidencioerrordeveloper: I'm running f32 and it has kernel-5.6.0-0.rc7.git0.2.fc32.x86_6419:04
errordeveloperok, thanks! I'll try that one then :)19:04
fidencioerrordeveloper: rawhide must have something similar19:05
errordeveloperI wonder who can unswer my main question though...19:05
fidencioerrordeveloper: the point is, CentOS kernel is way closer to RHEL kernel than to the Fedora kernel19:05
fidencioerrordeveloper: I'd suggest dropping an email to kata-dev, tbh19:05
errordeveloperfidencio: cheers, that would make sense!19:05
fidenciodevimc: just correct me if I'm mistaken here. In order to re-run Power8 CI job I can just do `/test-power`, right? Also, checking the docs I cannot see a simple way to re-trigger the centos-q35 test19:09
devimcfidencio: you're right19:10
devimcand there is no way to re-trigger centos-q3519:10
devimcfidencio: or you can try19:11
devimcjenkins.katacontainers.io/19:11
devimccan you login ?19:11
devimcI don't think so19:12
fidencioI don't think so :-)19:12
devimchah19:12
devimcok19:12
fidencioand I was wrong19:12
fidencioI can login there19:12
devimccool19:14
fidenciookay, I re-triggered that test19:14
devimcI guess you can retrigger the CI19:14
devimcokay, cool19:14
fidenciodevimc: would be worth to add shortcuts / commands to all the tests?19:14
fidenciodevimc: I mean, to all non-required tests19:14
devimcto all non-require CIs?19:15
devimcfidencio: you mean s/tests/CIs ?19:16
fidenciodevimc: yep19:16
devimcand shortcuts like /test-centos-q35 ?19:16
fidenciodevimc: yep19:16
devimcyep, that would be helpful19:17
devimcfidencio: i don't remember where is that documented19:18
devimcfidencio: do you have the link?19:18
fidenciodevimc: https://github.com/kata-containers/community/wiki/Controlling-the-CI19:18
devimcfidencio: thanks19:18
*** davidgiluk has quit IRC19:20
fidenciodevimc: IIUC, that would be a simple change on the jenkins side and then documenting the change19:20
devimcfidencio: yes, but I don't remember where xD19:22
devimcfidencio: you can ask Gaby19:22
devimcshe is the maintainer19:22
fidenciodevimc: sure, I'll bother her once we have the rhel-7-ci working19:23
fidenciodevimc: thanks!19:23
devimcfidencio: no problem19:24
*** jugs has quit IRC19:30
*** jugs has joined #kata-dev19:32
fidenciowell, at least I got consistent errors19:45
fidenciocrobinso: we don't use shimv2 on Fedora, do we?19:58
crobinsofidencio: no. this may be totally wrong but my understanding was shimv2 is a kube thing, which isn't applicable to Fedora podman usage19:59
fidenciocrobinso: seems that we may need that, as CRI-O has most of it implemented as well20:01
crobinsofidencio: yes back in a call a few months back dwalsh said cri-o can use shimv2 and it is an improvement. so it's something to look into for kube/openshift definitely20:02
crobinsofidencio: would be interesting if there's a way to validate it and work out the basic integration kinks outside of cri-o though but I don't know if that's an option20:03
*** igordc has joined #kata-dev20:27
*** ailan has quit IRC21:36
*** devimc has quit IRC22:15
*** sameo has quit IRC22:31
*** crobinso has quit IRC23:14
*** jugs has quit IRC23:21
*** jugs has joined #kata-dev23:21
*** dklyle has quit IRC23:55
*** david-lyle has joined #kata-dev23:55
*** david-lyle has quit IRC23:55
*** dklyle has joined #kata-dev23:57

Generated by irclog2html.py 2.15.3 by Marius Gedminas - find it at mg.pov.lt!