TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Ask HN: Why is eBPF a popular approach to monitoring?

6 点作者 jmartens超过 2 年前
For the developer that needs observability, why would they want an approach to monitoring that utilized eBPF instead of a language agent?

2 条评论

pushedx超过 2 年前
eBPF tracing programs are executed at tracepoints, kprobes, or uprobes, so they are entirely language agnostic, have access to the data structures only available to the kernel, persist across processes&#x2F;threads, are nearly guaranteed not to crash the system (by the verifier), and write to data structures that appear in well-known locations in the sysfs filesystem.<p>BSD and Solaris variants have had this type of thing for years with DTrace, which is “the way” to do observability on those systems. Now Linux has an even better way.<p>I highly encourage you to try out any of the tools scripts in the BCC repo. They’re just Python scripts that do really cool things that have never been done before! BPF is not that hard to get started with!<p><a href="https:&#x2F;&#x2F;github.com&#x2F;iovisor&#x2F;bcc&#x2F;tree&#x2F;master&#x2F;tools" rel="nofollow">https:&#x2F;&#x2F;github.com&#x2F;iovisor&#x2F;bcc&#x2F;tree&#x2F;master&#x2F;tools</a>
wmf超过 2 年前
For one thing eBPF is multi-language and it can see into the kernel.