Add Supervisor shutdown options #65
Conversation
|
Update: I discovered that using I've updated my PR to use I've also discovered if your actor traps exits, then you'll have to match on the following in your actor to shut it down, otherwise, it will timeout -> brutal kill. Down(ExitMessage(_supervisor_pid, Abnormal("Shutdown"))) -> {
actor.Stop(Abnormal("shutdown"))
} |
| ChildSpec(..child, shutdown: Timeout(timeout)) | ||
| } | ||
|
|
||
| pub fn shutdown_brutal_kill( |
There was a problem hiding this comment.
What's this function for? It is undocumented.
| ChildSpec(start: child.start, returning: updater, shutdown: child.shutdown) | ||
| } | ||
|
|
||
| /// Change the shutdown timeout |
There was a problem hiding this comment.
Could you add more documentation here please 🙏
| let result = | ||
| process.new_selector() | ||
| |> process.selecting_anything(fn(a) { a }) | ||
| |> process.select(0) |
There was a problem hiding this comment.
This drops the first message in the inbox, but this may not be the exit message. It needs to drop specifically the expected message here.
| // Ensure that the original processes are dead | ||
| should.be_false(process.is_alive(pid1)) | ||
| should.be_false(process.is_alive(pid2)) | ||
| should.be_false(process.is_alive(pid3)) |
There was a problem hiding this comment.
Could you move these to a new test please 🙏
The original issue that I was facing was that supervisors never actually killed their child processes, thus leading to a memory/process leak when my system was running for about a week.
This pull request implements the two strategies that erlang's supervisor employees, timeout and brutal kill.
The root cause of why actors never exit with erlang:exit(pid, normal) is still unknown.