Life of a Packet [I] - Michael Rubin, Google

Life of a Packet [I] - Michael Rubin, Google
Tracing the path of network traffic in the kubernetes system. Clarifying which API objects map to implementation and how Google deploys this in GKE today. Attendees will learn about topics from how networking packets are processed when the cluster is working as designed and what are common problems when the cluster is being creative and surprising.
About Michael Rubin
Senior Staff Eningeer & TLM, Google
Twenty years in the Systems Software Industry, from developing enterprise file servers and systems. The past ten years he has worked at Google where he founded the Linux Storage group for its data centers and worked on world wide WAN and BGP technologies. Today he is co-leading and managing Kubernetes with a focus on node, networking, storage and federation efforts.
Join us for KubeCon + CloudNativeCon in Barcelona May 20 - 23, Shanghai June 24 - 26, and San Diego November 18 - 21! Learn more at kubecon.io. The conference features presentations from developers and end users of Kubernetes, Prometheus, Envoy and all of the other CNCF-hosted projects.

[@GK-rl5du]

This has to be one of the timeless classic talks about Kubernetes Networking. Every one regardless of how mature the ecosystem gets, will/should hop on this amazing presentation.

[@musicmoodmaker]

That's the beauty of iptables

[@Indians-In-Germany-With-Dipak]

This is really awesome presentation ! love it . thank you !

[@mahmoudsaad1493]

Very informative talk, Thank you

[@rguntu9]

Good one.Thanks.

[@supericool]

super awesome talk for k8s network , never thought network can be this sexy.

[@joshuamiles8271]

Most excellent.

[@johnw.8782]

Really good talk. Thanks.

[@Devzx]

Thanks for this talk! I've been trying to find a comprehensive overview of networking (mainly how the different types of services work) and every blog I've read seems to have slightly different or even incorrect information.

[@charlesearle2055]

This is accurate! It's a challenge to learn k8s networking because of all the variations, misinformation and skipping-over-details to save time. I love this video, but an example of what I'm referring to is his example of pods-to-service where the traffic leaves the node but in the previous topic involved discussing overlay networks. Because it's already gone to IPtables does that handle any encapsulation? Or is it just assumed because we just talked about it that after IPtables gives you an endpoint destination outside the node does it now have to be delivered to the Overlay component for encapsulation?

[@acidas]

Thank you!

[@odedpriva]

great talk, any chance we can get the slide deck ?

[@amlivinginhell]

what is cbr0? what kind of bridge is it?

[@livno43615]

I think there's a mistake in minute 29, in the networkPolicy, you allowed the FE to send traffic to DB?

[@guibirow]

The Ins and Out of networking presentation: kzbin.info/www/bejne/r2PFmYluZrKZgbM

[@davidporter1190]

What the hell is the L2 discovery thing called here? “Our... L2” @6:13

[@tarunsharmaX]

I think he said find destination pod by ARPing and finding right dest mac/interface

[@grahamjerens5704]

The tee needs some fixing though

[@davidhope2199]

Lmao @ 1:12