Notes and exercises while reading through Programming Elixir by Dave Thomas.
imagine that instead your application consists of hundreds or thousands of processes, each handling just a small part of a request. If one of those crashes, everything else carries on. You might lose the work it’s doing, but you can design your applications to minimize even that risk. And when that process gets restarted, you’re back running at 100%.
In the Elixir and OTP worlds, supervisors perform all of this process monitoring and restarting.
Supervisor:
- manages one or more worker processes
- process that uses the OTP supervisor behavior
- via Erlang VM's process-linking and monitoring facilities
Add a supervisor to your stack application. Use iex to make sure it starts the server correctly. Use the server normally, and then crash it (try popping from an empty stack). Did it restart? What were the stack contents after the restart?
Notes:
Instead of using --sup flag and creating a new project, I added the start(_type, _args) function to the stack application. Also found I had to add mod: {Stack, []} to the application configuration.
When defining the child process in start, I initially defined it as:
worker(Stack.Server, [1, 2, 3])This resulted in an exception when attempting to start. The reason was that the list [1, 2, 3] is parsed as three arguments to Stack.Server.start_link and not as a single argument list. The correct definition is:
worker(Stack.Server, [ [1, 2, 3] ])- One approach is to use a separate worker process to store state.
- This process should be as simple and bullet proof as possible.
- This process is supervised separately - creating a supervision tree.
Supervision tree:
- Main Supervisor supervises the Stash Worker and a Subsupervisor
- Subsupervisor supervises our worker