I have some callback code that submits data over the net, gets an id
back and should write it to DB.
This takes a few seconds. I want to run it asynchronously, without
slowing down the request/response cycle. I would like to avoid a
worker/queue system if possible, to keep things simple.
I eventually want it in an observer, but I’m currently trying it out
right in a controller. If I do this:
def some_action
…
Thread.new {
Thread.abort_on_exception = true
logger.debug “!! 1”
sleep 5
logger.debug “!! 2”
sleep 5
logger.debug “!! Posts: #{Post.count}”
sleep 5
logger.debug “!! 3”
}
render :text => “foo”
end
Then visiting that page will log 1 and 2, but nothing else. 2 can log
after the action completes:
Completed in 4734ms (View: 4619, DB: 36) | 200 OK [http://
mysite.dev/]
!! 2
If the thread is this instead:
Thread.new {
Thread.abort_on_exception = true
logger.debug "!! 1"
logger.debug "!! 2"
logger.debug "!! Posts: #{Post.count}"
sleep 10
logger.debug "!! 3"
}
so Post.count runs before the action completes, then it works fine,
and 3 can output after the action completes:
…
!! Posts: 33
…
Completed in 5078ms (View: 4940, DB: 29) | 200 OK [http://
mysite.dev/]
!! 3
So basically, it seems that a thread in a controller action can run
after the action completes, but ActiveRecord can’t.
Is this an inherent limitation? Can I configure my way around it? I
was hoping 2.2 connection pools would mean this would work, but I
guess not.
I’ve looked at GitHub - imedo/background: deprecated - Please use successor project background_lite and
GitHub - tra/spawn: Placeholder for the repo previously known as spawn
but had issues with both, so I thought I’d try to figure it out with a
simple Thread for now.