Open Source Summit + Embedded Linux Conference + OSPOCon 2021

Fluent Bit has various configuration options for inputs, filters, parsers, and outputs plugins. Usually, these configuration options are put into one single file. It's not convenient if you just want to modify configurations for just one plugin especially if you want to modify it with Kubernetes API because the entire configuration file has to be loaded and parsed. Besides, Fluent Bit has to be restarted every time the configuration file is modified and this means the Pods of the entire DaemonSet have to be recreated in a K8s cluster. To address these concerns, the KubeSphere observability team initiated Fluent Bit Operator with which you can configure Fluent Bit with CRDs like FluentBit, FluentBitConfig, Input, Filter, Parser, Output in a Kubernetes-native way. What's more, Fluent Bit DaemonSet Pods will not be recreated every time a config changes. It supports dynamic configuration and provides great flexibility in building the Kubernetes-native logging layer. DigitalOcean adopts Fluent Bit Operator as the Kubernetes log processor and made great contributions to Fluent Bit Operator. In this talk, we will introduce the Fluent Bit Operator and its architecture, then deep dive into the KubeSphere logging system based on Fluent Bit Operator.