[KubeCon China 2020]BPF for Chaos and Tracing in Kubernetes(in Chinese)

BPF for Chaos and Tracing in Kubernetes – Wenbo Zhang, PingCAP

In the world of distributed computing, faults and performance problems happen to clusters unpredictably, especially when they run in cloud. Among all the faults types, kernel failures are perhaps the hardest to analyze and simulate. A practical solution for this is Berkeley Packet Filter (BPF), a highly flexible and efficient virtual machine-like construct in the Linux kernel that allows bytecode execution at various hook points in a safe manner.

In this session, Wenbo Zhang will share his BPF practices on a distributed and transactional key-value database known as TiKV. He will introduce what BPF is, how to develop BPF programs to analyze kernel-caused performance problems, and how to do kernel chaos experiments on Kubernetes. He will also talk about how the team converted the BCC-based BPF application to make “Compile Once – Run Everywhere“ come true on the 5.x kernel.

Follow TiKV on Twitter https://twitter.com/tikvproject

source by TiKV

linux foundation