An introduction to control groups (cgroups) version 2 - Michael Kerrisk - NDC TechTown 2021

Control groups (cgroups) allow us to limit and measure the resources used by groups of process. They are one of the fundamental building blocks of container frameworks, as well as a number of other interesting tools such as Flatpak and Firejail. We have (finally) reached the point where cgroups version 2 is supplanting cgroups version 1, as most major Linux distributions make the switch.
In this presentation, we will look at how cgroups work, from the perspective of the future. That is to say, I will ignore cgroups v1, and explore how cgroups work starting fresh from a version 2 perspective. Topics we’ll cover include creating and destroying cgroups, moving processes into cgroups, setting resource limits on cgroups, enabling and disabling controllers, and managing different resources to different levels of granularity. No previous knowledge of cgroups will be assumed.

[@FrederikoCosta]

This talk, by itself, is worth 10x more than any Aws reinvent talk.

[@victoryosikwemhe2280]

100%

[@crikxouba]

And for the price of one reinvent ticket talk, you can pay for one of his courses, which would be so much more useful.

[@chainer22]

Probably one of the clearest explanation of cgroup out there. Thank you!

[@TheKarantan]

My thoughts exactly. Also, I loved the demos.

[@tommasoborgato]

Great cgroups V2 explanation! Couldn't find anything better

[@kgsrivastava]

Excellent video sir. It's the best I have seen on cgroups. The way you explain is awesome. The pace at which you speak is not rushed like other videos and hence easy to understand.

[@jzburns1990]

Fantastic talk, fantastic resources from Michaels homepage

[@dassyareg7587]

This is just awesome. Great job, Michael.

[@Devzx]

Amazing talk, really helped solidify how cgroups work! Thank you!

[@victoryosikwemhe2280]

Thank you so much for this. It is really helpful to understand how this works

[@MichaelKerrisk]

See also the follow-on presentation at ruclips.net/video/Clr_MQwaJtA/видео.html

[@jonnytheponny5753]

Thanks so much for the absolutely great and informative Talk, this talk (Part 1) is amazing and so helpful. Thanks again!

[@Shri]

Thank you Sir!

[@hardas81]

Excellent presentation! Thank you so much! Not only did you explain everything perfectly, you managed to spark my curiosity to learn more. See you on the next video :)

[@Shri]

Brilliant talk!

[@uberwebd9824]

I love the way you teach - thank you!

[@hardikrajpal2410]

Any idea where we can get the cpu_burner program used at 42:48?

[@uberwebd9824]

at 19:36 shouldn't it have blcoked after 3 sleeps ? (for loop + shell + 3 sleeps = 5 procs)

[@christianburke4220]

If echo is built in, I imagine a construct like for is built in as well.
I wonder if that's the reason for the apparently disparate styles on for loops in different shells? They are not standard utilities?

[@MichaelKerrisk]

@@christianburke4220 Yes, your explanation is correct. 'for' is a shell built-in, and so does not result in a separate process.

[@pixelPlex]

Does the sub cgroup thing ( ruclips.net/video/kcnFQgg9ToY/видео.html ) work with the posix_spawn function?

[@karatebreakfast6908]

47:56 ".....But why male models?"

[@daniel_andrade]

In bash we can do `$ echo $$ | sudo tee /sys/fs/cgroup/grp1/cgroup.procs` then anything run on that shell will execute in grp1.