| *** sameo has quit IRC | 02:23 | |
| kata-irc-bot | <dgibson> @archana.m.shinde @julio.montes so, I'm less interested in what new versions we support as whether we can drop support for old qemu versions (e.g. 2.10) | 03:23 |
|---|---|---|
| *** bpradipt has joined #kata-dev | 04:21 | |
| *** pcaruana has joined #kata-dev | 06:20 | |
| kata-irc-bot | <archana.m.shinde> @dgibson Yes, I think we can drop support for qemu 2.10 especially if it helps simplify code | 06:21 |
| *** dklyle has quit IRC | 06:26 | |
| *** jodh has joined #kata-dev | 06:45 | |
| kata-irc-bot | <dgibson> @archana.m.shinde thanks, good to know | 06:50 |
| kata-irc-bot | <dgibson> @archana.m.shinde and the next question is just how many can we drop | 06:50 |
| kata-irc-bot | <dgibson> can we drop support for 3.1? 4.0? 4.1? | 06:50 |
| kata-irc-bot | <dgibson> not 4.1, I guess | 06:50 |
| *** sameo has joined #kata-dev | 07:17 | |
| *** sameo has quit IRC | 07:20 | |
| *** sameo has joined #kata-dev | 07:20 | |
| *** hashar has joined #kata-dev | 07:26 | |
| *** davidgiluk has joined #kata-dev | 08:02 | |
| *** bpradipt has quit IRC | 08:09 | |
| *** hashar has quit IRC | 08:16 | |
| *** hashar_ has joined #kata-dev | 08:16 | |
| *** hashar_ is now known as hashar | 08:25 | |
| *** bpradipt has joined #kata-dev | 09:23 | |
| *** hashar is now known as hasharAway | 09:31 | |
| *** amorenoz_ has quit IRC | 10:07 | |
| *** amorenoz has joined #kata-dev | 10:10 | |
| *** hasharAway is now known as hasharLunch | 10:16 | |
| *** bpradipt has quit IRC | 10:38 | |
| *** bpradipt has joined #kata-dev | 10:43 | |
| *** amorenoz has quit IRC | 11:18 | |
| *** amorenoz_ has joined #kata-dev | 11:18 | |
| *** devimc has joined #kata-dev | 11:48 | |
| *** hasharLunch is now known as hashar | 11:51 | |
| fidencio | devimc: hey, I'm not sure what to do with the forward ports I've posted | 12:16 |
| fidencio | devimc: travis doesn't seem to work for me :-/ | 12:16 |
| devimc | fidencio: lemme take a look | 12:30 |
| devimc | fidencio: https://github.com/kata-containers/kata-containers/pull/310 ? | 12:31 |
| fidencio | devimc: that one as well, but mainly: https://github.com/kata-containers/kata-containers/pull/304 | 12:32 |
| devimc | fidencio: done | 12:34 |
| devimc | xD | 12:34 |
| devimc | https://travis-ci.org/github/kata-containers/kata-containers/builds/701250122?utm_source=github_status&utm_medium=notification | 12:34 |
| fidencio | devimc: if you have a hammer, every problem becomes a nail | 12:35 |
| devimc | haha | 12:35 |
| *** georgk00 has quit IRC | 12:52 | |
| *** georgk has joined #kata-dev | 12:52 | |
| *** dklyle has joined #kata-dev | 12:58 | |
| *** devimc has quit IRC | 13:57 | |
| *** devimc has joined #kata-dev | 13:58 | |
| fidencio | devimc: can you hammer this one as well? https://github.com/kata-containers/kata-containers/pull/310 I've tried Yesterday, without success | 14:12 |
| devimc | fidencio: sure, I can try | 14:13 |
| fidencio | devimc: thanks! | 14:14 |
| devimc | yey it works | 14:14 |
| *** amorenoz__ has joined #kata-dev | 14:41 | |
| *** amorenoz_ has quit IRC | 14:43 | |
| *** amorenoz__ has quit IRC | 14:43 | |
| *** amorenoz has joined #kata-dev | 14:45 | |
| *** amorenoz has quit IRC | 14:48 | |
| *** amorenoz_ has joined #kata-dev | 14:48 | |
| *** amorenoz_ has quit IRC | 14:48 | |
| *** amorenoz has joined #kata-dev | 14:49 | |
| *** amorenoz has quit IRC | 14:53 | |
| *** amorenoz has joined #kata-dev | 14:54 | |
| *** crobinso has joined #kata-dev | 15:04 | |
| *** sameo has quit IRC | 15:18 | |
| *** sameo has joined #kata-dev | 15:19 | |
| kata-irc-bot | <eric.ernst> devimc: hey there | 15:36 |
| kata-irc-bot | <eric.ernst> i believe containerd/cgroups has v2/rootless now | 15:36 |
| *** amorenoz_ has joined #kata-dev | 15:45 | |
| *** amorenoz has quit IRC | 15:48 | |
| devimc | hey Eric, that's really cool | 15:53 |
| kata-irc-bot | <fidencio> @gabriela.cervantes.te, @salvador.fuentes, would you be able to test a patch for CRI-O that may fix the issue you pointed out in the meeting? | 15:55 |
| kata-irc-bot | <salvador.fuentes> @fidencio sure thing | 15:56 |
| kata-irc-bot | <eric.ernst> based on that, devimc, i’m wondering if we can cgroup mgr. | 15:57 |
| kata-irc-bot | <eric.ernst> It’s kind of a pain, since it works at a ‘different level’ than we use elsewhere (ie, we seem to work more with LinuxResources, where this is working with their own cgroup data structure). | 15:57 |
| kata-irc-bot | <eric.ernst> Rather than translate and keep state around for the cgroups, I think the container/cgroups was a bit easier? | 15:58 |
| kata-irc-bot | <eric.ernst> were there any other reasons for using libcontainer’s cgroup? | 15:58 |
| kata-irc-bot | <fidencio> https://github.com/fidencio/cri-o/tree/wip/pass-runtime-type-to-the-runtimes-option | 15:59 |
| kata-irc-bot | <fidencio> Using this branch you should be able to pass the runtime type as VM | 15:59 |
| devimc | @eric.ernst we already have cgroup mgr to encapsulate this, so the transition to containerd/cgroups should be easy, we move to libcontainer because it supports v2 and partially rootless | 16:00 |
| kata-irc-bot | <eric.ernst> I wonder if we need the encapsulation / what it buys us. | 16:05 |
| kata-irc-bot | <eric.ernst> Avoid need to ’load cgroup?” | 16:05 |
| *** jodh has quit IRC | 16:05 | |
| kata-irc-bot | <eric.ernst> and, that currently doesn’t really handle updating the cgroups based on new resources. | 16:06 |
| kata-irc-bot | <eric.ernst> anyway, i’ll look at it more, but I think it may make sense to move away from the libcontainer usage and just use containerd/cgroups | 16:07 |
| kata-irc-bot | <salvador.fuentes> ok thanks, let me try | 16:14 |
| kata-irc-bot | <fidencio> let me know if it works and I'll submit the PR to cri-o, if it does | 16:36 |
| kata-irc-bot | <salvador.fuentes> yeah, that worked, but now I am getting other errors, but not related to this issue: ``` # time="2020-06-23 16:36:34.229482427Z" level=debug msg="Response error: failed to create pod network sandbox k8s_podsandbox1_redhat.test.crio_redhat-test-crio_1(ca12146bf1a3bffbee6a8677755e89068c436b0cebc54e931 6a94b9d43b0812b): error adding loopback interface: failed to Statfs \"\": no such file or directory" | 16:43 |
| kata-irc-bot | file="go-grpc-middleware/chain.go:25" id=165cf8c4-3b45-484e-a0bb-bc374d577580 name=/runtime.v1alpha2.RuntimeSe rvice/RunPodSandbox # time="2020-06-23T16:36:34Z" level=fatal msg="run pod sandbox failed: rpc error: code = Unknown desc = failed to create pod network sandbox k8s_podsandbox1_redhat.test.crio_redhat-test-crio_1(ca12146bf1a3bffbee 6a8677755e89068c436b0cebc54e9316a94b9d43b0812b): error adding loopback interface: failed to Statfs \"\": no | 16:43 |
| kata-irc-bot | such file or directory"``` | 16:43 |
| kata-irc-bot | <salvador.fuentes> @fidencio I think you can open the PR to fix the runtime_type issue, in the meantime I'll check this other issue | 16:46 |
| kata-irc-bot | <fidencio> Okay, my I take a look at the config used? | 16:47 |
| kata-irc-bot | <salvador.fuentes> I think it is the default one, iirc, the tests do not take the config in /etc/crio/crio.conf, right? I only exported RUNTIME=containerd-shim-kata-v2 and RUNTIME_TYPE="vm" | 16:49 |
| kata-irc-bot | <eric.ernst> or, wdyt re: cgroup manager? | 16:58 |
| kata-irc-bot | <eric.ernst> it made sense to me when we were using the libcontainer cgroup manager, but not sure now. | 16:58 |
| kata-irc-bot | <eric.ernst> Or, devimc, can you clarify the gap with existing solution? | 17:00 |
| kata-irc-bot | <eric.ernst> (ie, device support?) | 17:00 |
| kata-irc-bot | <fidencio> would be good to also export `RUNTIME_ROOT="/run/vc"` | 17:05 |
| devimc | @eric.ernst rootless cgroups, not sure if libcontainer already supports them | 17:06 |
| kata-irc-bot | <fidencio> Also, is there some way to export``manage_ns_lifecycle = true ?` | 17:06 |
| kata-irc-bot | <eric.ernst> ok. Do we have what we need wrt containerd/cgroups for device support? | 17:12 |
| kata-irc-bot | <eric.ernst> FWICS, using *just* containerd/cgroups may be feasible at this point; it looks like cgroup manager pkg you made is mostly using libcontainer, and seems to be mostly focused on devices | 17:13 |
| *** hashar has quit IRC | 17:25 | |
| *** hashar has joined #kata-dev | 17:25 | |
| kata-irc-bot | <salvador.fuentes> oh ok, not sure about `manage_ns_lifecycle` , let me check | 17:31 |
| kata-irc-bot | <fidencio> Also, seems that my patch screwed up setting runtime_root :slightly_smiling_face: | 17:35 |
| kata-irc-bot | <fidencio> I'm cooking a new version here | 17:35 |
| kata-irc-bot | <salvador.fuentes> ohh ok, thanks | 17:36 |
| kata-irc-bot | <fidencio> Okay, I've force-pushed the patch to the same branch | 17:50 |
| kata-irc-bot | <fidencio> That one should be good to go | 17:50 |
| kata-irc-bot | <salvador.fuentes> thanks, checking | 17:52 |
| *** bpradipt has quit IRC | 18:18 | |
| *** hashar is now known as hasharAway | 18:23 | |
| kata-irc-bot | <eric.ernst> devimc ^^ | 19:01 |
| devimc | @eric.ernst we can reuse that package, just the replace the calls to libcontainer with calls to containerd/cgroups | 19:07 |
| devimc | the device cgroup was not easy to support | 19:07 |
| *** davidgiluk has quit IRC | 19:22 | |
| *** Yarboa has joined #kata-dev | 19:34 | |
| kata-irc-bot | <salvador.fuentes> exported those variables and updated with your latest changes. Now I am getting ```# time="2020-06-23T20:11:35.889232355Z" level=error msg="createContainer failed" ID=19e5a55f6f4ef161acfef29886d1a3cf183e55f76788e25a9d0b67b26787e424 error="rpc error: code = Internal desc = EINVAL: Invalid argu$ ent" source=virtcontainers subsystem=kata_agent``` | 20:14 |
| kata-irc-bot | <salvador.fuentes> that comes from kata | 20:14 |
| kata-irc-bot | <salvador.fuentes> any idea @archana.m.shinde ^ | 20:14 |
| kata-irc-bot | <eric.ernst> The main item I was questioning if it makes sense to keep an entire structure in place. | 20:19 |
| kata-irc-bot | <eric.ernst> It seems we would need to recalculate most of it (linux resources) each time there’s a container update. | 20:20 |
| kata-irc-bot | <eric.ernst> might be good to talk through in realtime. | 20:46 |
| *** devimc has quit IRC | 20:51 | |
| kata-irc-bot | <archana.m.shinde> But looks like that could be from the agent | 21:02 |
| kata-irc-bot | <archana.m.shinde> can you enable kata agent logs and see if you find a more useful message? | 21:02 |
| *** sameo has quit IRC | 21:06 | |
| kata-irc-bot | <salvador.fuentes> @archana.m.shinde seems to be related to cgroup manager. As default it is configured as systemd and thats the error I got. Tried to change to cgroupfs and now I get: | 21:17 |
| kata-irc-bot | <salvador.fuentes> ```# time="2020-06-23 20:57:38.407141596Z" level=debug msg="Response error: cri-o configured with cgroupfs cgroup manager, but received systemd slice as parent: pod_123-456.slice" file="go-grpc-middleware/chain.go:25" id=5894d1d4-a704-4e65-80c0-b34a5ec8f5b9 name=/runtime.v1alpha2.RuntimeService/RunPodSandbox``` | 21:17 |
| kata-irc-bot | <salvador.fuentes> seems like somewhere else I need to change the configuration? | 21:17 |
| kata-irc-bot | <archana.m.shinde> yeah | 21:17 |
| kata-irc-bot | <salvador.fuentes> btw, for kata 1.x I do not have to change all this, is this different on both implementations? | 21:17 |
| kata-irc-bot | <archana.m.shinde> maybe it is configured in the tests somewhere | 21:18 |
| kata-irc-bot | <archana.m.shinde> fuentess: not sure | 21:23 |
| kata-irc-bot | <archana.m.shinde> https://github.com/fidencio/cri-o/blob/23193ea43643053d4b118daccd2d8c0e336f455b/test/helpers.bash#L68 | 21:23 |
| kata-irc-bot | <archana.m.shinde> but I see the above that the cgroup manager is systemd there | 21:23 |
| kata-irc-bot | <archana.m.shinde> should check with devimc if some cgroup support is missing for shimv2 although I doubt it | 21:24 |
| kata-irc-bot | <fidencio> I've been using shimv2 with cgroupsv1, haven't faced issues strict related to that | 21:25 |
| kata-irc-bot | <fidencio> @salvador.fuentes, what's the easiest way for me to try that? | 21:25 |
| kata-irc-bot | <salvador.fuentes> yeah, it is weird as I can also run k8s | 21:25 |
| kata-irc-bot | <salvador.fuentes> so if you already have an environment with kata-shimv2 and you have the tests repo cloned, you can try with these:... | 21:26 |
| kata-irc-bot | <fidencio> I have an environment with shimv2 setup, but I'm not sure how easily I can build something in that machine (RHCOS, immutable OS, joy ...) But, gimme the instructions and I'll give it a try Tomorrow (too late to connect back to the VPN Today ;-)) | 21:27 |
| kata-irc-bot | <salvador.fuentes> ```export CONTAINER_CGROUP_MANAGER=cgroupfs export RUNTIME_ROOT=/run/vc export RUNTIME_TYPE=vm export RUNTIME=containerd-shim-kata-v2 # From tests repo: sudo -E PATH=$PATH make crio``` | 21:28 |
| kata-irc-bot | <salvador.fuentes> and of course, build cri-o with your changes | 21:28 |
| kata-irc-bot | <salvador.fuentes> that is what I have tried locally | 21:28 |
| kata-irc-bot | <fidencio> right, I'll give it a try Tomorrow and let you know! | 21:28 |
| kata-irc-bot | <salvador.fuentes> ok, thanks Fabiano :slightly_smiling_face: | 21:29 |
| kata-irc-bot | <fidencio> np! and the atch adding the RUNTIME_TYPE should be merged soon, already got the needed reviews, it's just a matter of a clean CI run | 21:29 |
| kata-irc-bot | <salvador.fuentes> great, thanks | 21:30 |
| kata-irc-bot | <archana.m.shinde> fuentess: Wonder if it is picking any configs from /etc/crio/crio.conf | 21:31 |
| kata-irc-bot | <archana.m.shinde> I remember it has a cgroup config there as well | 21:32 |
| kata-irc-bot | <salvador.fuentes> I already looked at that configuration and I have: `cgroup_manager = "cgroupfs"` and `conmon_cgroup = "pod"` , although not sure if both settings are related? | 21:33 |
| kata-irc-bot | <fidencio> @salvador.fuentes what does your `/etc/crio/crio.conf` looks like? is there something in `/etc/crio/crio.conf.d`? | 21:34 |
| kata-irc-bot | <fidencio> and the kata configuration file? | 21:34 |
| kata-irc-bot | <salvador.fuentes> @fidencio this is my crio.conf: https://seashells.io/v/gaQPuRND and nothing inside /etc/crio/crio.conf.d | 21:39 |
| kata-irc-bot | <salvador.fuentes> and on Kata, I don't have any special configuration, just default ones and enabled logging: https://seashells.io/v/3v7gjPcZ | 21:40 |
| *** Yarboa has quit IRC | 21:43 | |
| kata-irc-bot | <fidencio> not exactly comparing side-by-side but one thing that I could spot here | 21:49 |
| kata-irc-bot | <fidencio> I use: `sanbox_cgroup_only=true` | 21:50 |
| kata-irc-bot | <fidencio> I also use q35 as machine type | 21:50 |
| kata-irc-bot | <fidencio> and both my guest and host do *not* have cgroupsv2 enabled | 21:50 |
| *** hasharAway has quit IRC | 22:11 | |
| *** th0din has quit IRC | 23:15 | |
| *** th0din has joined #kata-dev | 23:16 | |
| *** kgz has quit IRC | 23:29 | |
| *** kgz has joined #kata-dev | 23:31 | |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!