Designing for Scalability with Erlang/OTP

Chapter 5. Controlling OTP Behaviors

We have in the previous chapters covered the highlights of the gen_server behavior. You should by now have implemented your first client-server application and started to build an idea of how OTP behaviors help you to reduce your code base by allowing you to focus on the specifics of what your system has to do. This chapter digs deeper into behaviors, exploring some of the advanced topics intermixed with built-in functionality. While we are focusing on generic servers, most of what we write will apply to many of the other behaviors, including those you could implement yourself. Read with care, as we reference this chapter often in the remainder of this book.

The sys Module

We’ve mentioned many times the built-in functionality you get as a result of using OTP behaviors and the ease with which you can add your own features. Most of what we cover is accessed through the sys module, allowing you to generate trace events, inspect and manipulate behavior state, as well as send and receive system messages. All of this functionality works on the standard OTP behaviors, but also, as we show in Chapter 10, you can reuse it when defining your own behaviors.

Tracing and Logging

Let’s find out how built-in tracing works by running a little example. Start your frequency server in the shell and, using the sys module, try the following:

1> frequency:start().
{ok,<0.35.0>}
2> sys:trace(frequency, true).
ok
3> frequency:allocate(). *DBG* frequency got call {allocate,<0.33.0>} ...