| kata-irc-bot | <jakob.naucke> All the scripts or are you specifically thinking about any scripts in particular? I think Bash would work better e.g. for `tests/.ci/` stuff | 07:47 |
|---|---|---|
| kata-irc-bot | <samuel.ortiz> @julio.montes Good morning | 13:48 |
| devimc | Hi Samuel, good morning | 14:02 |
| kata-irc-bot | <samuel.ortiz> @julio.montes I’m looking at the host cgroups code for the runtime, when `sandbox_cgroup_only` is not set. It seems to me that the documentation (https://github.com/kata-containers/documentation/blob/master/design/host-cgroups.md) is not aligned with the code: IIUC we’re creating a new cgroup for each container (and each sandbox), and adding the shim process to that cgroup. And the only limit that we’re applying are CPU ones, | 14:13 |
| kata-irc-bot | memory. Does that sound accurate? | 14:13 |
| devimc | yes, that's right | 14:23 |
| kata-irc-bot | <samuel.ortiz> @julio.montes So, e.g. `Move hypervisor and applicable threads to memory cgroup /kata` is incorrect? | 14:26 |
| kata-irc-bot | <samuel.ortiz> @julio.montes Or things like: `Per each container, move its kata-shim to its own containerCgroupHost` is also incorrect and applies to kata 1.x but not 2.x ? | 14:27 |
| kata-irc-bot | <samuel.ortiz> @julio.montes Just checking if my understanding is right :slightly_smiling_face: | 14:27 |
| devimc | "Move hypervisor and applicable threads to memory cgroup /kata" is correct | 14:28 |
| devimc | we move vcpus (threads) into cgroups with constraints. | 14:30 |
| kata-irc-bot | <samuel.ortiz> > “Move hypervisor and applicable threads to memory cgroup /kata” is correct | 14:30 |
| kata-irc-bot | <samuel.ortiz> > we move vcpus (threads) into cgroups with constraints. With `sandbox_cgroup_only` is set to false? | 14:33 |
| kata-irc-bot | <samuel.ortiz> @julio.montes ^ | 14:34 |
| devimc | yes | 14:47 |
| devimc | @samuel.ortiz | 14:47 |
| kata-irc-bot | <samuel.ortiz> @julio.montes Could you please point out where in the code that’s done? | 14:48 |
| devimc | sanbox.go:2080 | 14:48 |
| kata-irc-bot | <samuel.ortiz> @julio.montes So `constrainHypervisor` is only called when `s.state.CgrouPath` is set, and afaict `s.state.CgroupPath` is set from `setupSandboxCgroup` The latter is only called from `createSandboxFromConfig` iff `s.Config.SandboxCgroupOnly` is true. That’s my analysis, but I may be wrong. | 14:58 |
| kata-irc-bot | <samuel.ortiz> What am I missing here? | 14:59 |
| kata-irc-bot | <samuel.ortiz> @julio.montes As I’m willing to fix this code if necessary, is there anything I’m missing here? | 15:52 |
| devimc | @samuel.ortiz after reading the code, I think you are right, something changed (bug was introduced ..?) and now vcpus threads are not constrained :| | 15:57 |
| kata-irc-bot | <wmoschet> @jakob.naucke I would like to improve the way CI jobs can be reproduced locally. I was thinking in write new code in Python, if yet another language is acceptable | 16:39 |
| kata-irc-bot | <jakob.naucke> like `tests/integration/` and whatnot? | 16:44 |
| kata-irc-bot | <salvador.fuentes> @samuel.ortiz, devimc, seems like the irc-bot not working correctly as I do not see here (slack) last devimc's response. So pasting here: [10:57:26] <devimc> @samuel.ortiz after reading the code, I think you are right, something changed (bug was introduced ..?) and now vcpus threads are not constrained :| | 17:08 |
| kata-irc-bot | <wmoschet> In a nutshell, I want to have a make target (e.g. `make job-cri_containerd_k8s`) which will run the Jenkins's `cri_containerd_k8s` job in a vm in my localhost. It is the extension of this work: https://github.com/kata-containers/tests/pull/3547 | 18:50 |
| kata-irc-bot | <samuel.ortiz> @salvador.fuentes Thanks! I’ll open an issue for that bug, I wanted to rework a bit how we handle host cgroups anyways. | 21:02 |
| kata-irc-bot | <samuel.ortiz> @salvador.fuentes @julio.montes and I’ll get back to IRC as well ;) | 21:02 |
Generated by irclog2html.py 2.17.2 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!