Kindling
Overview and Concepts
Overview
Trace Profiling
RPC Trace
Roadmap
Prometheus vs. Kindling vs. APM
How Kindling Agent is going to evolve
Installation
Kindling Agent
Requirements
Install Kindling in Kubernetes
Setting up Grafana
FAQ
Download Linux kernel headers
Usage
How to enable Trace Profiling
Prometheus Metrics Description
Use Cases
Service Map and Performance
Observe Java Lock
Developer Guide
Architecture
Build Kindling container image from source codes
Kindling agent-libs 用户态空间数据流
Kindling agent-libs内核态空间数据流
Kindling collector 流水线数据传递流程
Add new eBPF hooks
Kindling probe核心流程
Developer FAQ
Reference
Kindling Java Agent
Overview
Modifications of async-profiler
async-profiler 改造点
Trace Profiling Operation Manual
本文档使用 MrDoc 发布
-
+
home page
Architecture
## The Project Architecture  ## Agent Architecture Kindling agent from detailed view can be splited into **Kindling Probe** and **Kinding Collector**. - **Kindling Probe** is composed of kernel-space modules which produce **kernel events** and **user-space** module controller which consumes **kernel events** and transform them into structured format, namely **Kindling Events.** - **Kindling Collector** process and analyze **Kindling Events** into metrics and traces with the kubernetes metadata, and export them to the Prometheus server. ## Detailed Data Flow The data flows with the following sequence: 1. **Kernel Events** are collected from kernel space, passed through ringbuffer: eBPF module uses perf-buffer, while kernel module uses ringbuffer self-defined. 2. In user-space, **Kindling Probe** correlates container information, thread information, fd information to **Kernel Events** (if it has, some events may fail to correlate to any thread or fd). 3. **Kindling Events** will be transferred from **Kindling Probe** to **Kindling Collector** by **Event Publisher** through _CGO_. 4. Then **Kindling Collector** will parse **Kindling Events**, judge if it's a request or reponse, analyze its protocol, etc. **Collector** also aggregates requests data for sending and integrating kubernetes metadata, and finally export the yields (metrics and traces) to the Prometheus server.
xieyun
Oct. 26, 2022, 4:15 p.m.
Share documents
Collection documents
Last
Next
Scan wechat
Copy link
Scan your mobile phone to share
Copy link
关于 MrDoc
觅思文档MrDoc
是
州的先生
开发并开源的在线文档系统,其适合作为个人和小型团队的云笔记、文档和知识库管理工具。
如果觅思文档给你或你的团队带来了帮助,欢迎对作者进行一些打赏捐助,这将有力支持作者持续投入精力更新和维护觅思文档,感谢你的捐助!
>>>捐助鸣谢列表
微信
支付宝
QQ
PayPal
Markdown文件
share
link
type
password
Update password