when its just about downtime, then you can keep the current deployment active, while the next one is building and after the building part is done, it automatically will be promoted to the current deployment

But I need to restart, no?

after you set it up properly, you dont need to. I guess you are self hosting, right? Write a deployment script, that does all of this for you and you have 0.0 downtime caused through builds. If you are on plattforms like vercel or using self hosting "helpers" like coolify, it will be handled automaticallyfor you

Yes I am selfhosting using pterodactyl

It‘s running in a turborepo and building everything on restart

great. Instead of restarting the whole server, just restart the specific container (that then builds and deploys)

Yes but that causes the server to go down

normally you have one server running and it continues to stay online and then just your docker containers on that machine restarts, not that the whole server restarts. Thats also a reason why we have docker

Mo I don‘t restart the entire server

My bad

I mean the container

When I want to update the website, I need to restart the container

I cannot build it before restarting

Its just

Click start → build → container started → stop → start → build → container started

And that building takes like a min

Which cuases a minute of downtime

ah now it makes more sense yea. You can create a building container for your app or a deployment script, that builds the whole thing and after the build it finished, it takes down the old container and instantly starts it, so there is no downtime.

As we working with containers, you can also have multiple containers open and restart one after another, so there is 100% no downtime

Depends on how you host the application. If you are using Docker maybe you can take inspiration from a setup I have in production.

I host the Next.js container on a small VPS which is not really powerful and building the image takes ~8 minutes on it. Because of this I created a script which offloads the building of the image to my local machine where it takes ~1.5 minutes. After the image is built I compress it to a image.tar file, and send it to VPS using ssh and scp.

Once the image is uploaded to the VPS I swap the production image using docker load and docker compose to restart the container with the new image.

This ensures that I have about <1 second downtime, while not having to do some complicated blue-green setup which slows down the production server.