Add timeout to gunicorn in heroku

[mvandenburgh]

Downstreaming kitware-resonant/cookiecutter-resonant#190 from the cookiecutter.

This will help distinguish Heroku request timeouts from other errors that cause a SystemExit in sentry.

[waxlamp]

What's a good way to test this? Maybe an endpoint with an artificial delay that we can introduce and then remove later?

[mvandenburgh]

What's a good way to test this? Maybe an endpoint with an artificial delay that we can introduce and then remove later?

Yeah I think that's the only way we could reliably test this. We could add a time.sleep call to an endpoint and test in staging.

[waxlamp]

What's a good way to test this? Maybe an endpoint with an artificial delay that we can introduce and then remove later?

Yeah I think that's the only way we could reliably test this. We could add a time.sleep call to an endpoint and test in staging.

Can you file an issue and assign it to yourself?

[jjnesbitt]

FWIW I think making a temporary endpoint and merging it to master to test staging is the wrong approach. You can create this endpoint and run gunicorn locally, which imo is sufficient.

[mvandenburgh]

Oh yeah i think you're right. For some reason I thought we had to test this in Heroku because we're testing the Heroku timeout, but in reality we are replacing the Heroku 30-second timeout with a 25-second timeout built in to gunicorn

I'll test it locally with gunicorn, @waxlamp does that sound good?

[waxlamp]

Yes, good idea.

[dandibot]

🚀 PR was released in v0.2.26 🚀