Build Stuff 2016 Workshops

In this workshop we will explore the exciting world of modern Linux tracing tools. This is not another workshop that shows you examples of running `top` followed by `vmstat`. We will drill down and discuss various tracing features built into the system: the venerable ftrace, which provides access to kernel tracepoints, kprobes, and uprobes; perf_events and the perf front-end, which also offers support for low-level CPU events; a glimpse at the (declining).

SystemTap framework; and finally, the icing on the cake, BPF and the BPF Compiler Collection (BCC), which has support for kprobes, uprobes, tracepoints, and USDT probes (such as those in Node and the JVM). We will explore a series of demos covering a variety of tracing scenarios -- from block I/O and the network stack to kernel memory leaks and high CPU utilization in user-space processes. We will also see how to build our own tracing tools from scratch, and how to run them in production continuously. During the workshop you will perform numerous short labs: if you come from a development background, you will use some existing tools and also write your own; if you come from an operations background, you will use a variety of tools and extend them as applicable for your own scenarios.