Hello, By default, when a new deployment goes into the `ACTIVE` state we will send `SIGTERM` to the old container, wait 3 seconds for it to gracefully shut down, and if it hasn't shut itself down after the 3 seconds, we force kill it. You can control that `SIGTERM` to `SIGKILL` time with a `RAILWAY_DEPLOYMENT_DRAINING_SECONDS` variable - <https://docs.railway.com/reference/deployments#singleton-deploys> So once you extend that time frame, any new incoming requests HTTP requests would be routed to the new deployment, while we wait for the old deployment to either gracefully shut down, or for whatever time you have set until we kill it. Hope this helps! Best, Brody

Ok, thanks. And if I change RAILWAY\_DEPLOYMENT\_DRAINING\_SECONDS to 600 seconds, for example, that means that if the SIGTERM signal does not stop the container, we wait 10 minutes before forcing the container to stop. Is that correct? And do all new requests that arrive in the 10-minute interval go to the old or new deployment?

Correct. New deployment.