Someone has to patch the web server. On serverless it is not you.

A flaw in a web server or a kernel turns somebody's evening into a maintenance window: find the machines, wait for the package, restart the right things in the right order, reboot. Which of that is yours is decided by how you host, not by how careful you are.

Who patches what

The web server and the kernel are part of the machine, not of your code. Whoever runs the machine installs the updates, and whoever installs them plans the restart.

How you host The web server The kernel
A VPS or dedicated server Yours: a package from your distribution, then restart the service Yours: a package, then a reboot
Docker on servers you run Yours: rebuild the image, restart every container Yours, on every host, with a reboot
A cluster you run Yours, the ingress controller included Yours, on every node, one at a time
Serverless containers Only what your own image holds The provider's

The row you are on decides how the next advisory reads: a list of machines to work through, or a paragraph you read and move on from.

What the work is

Patching is not one action, and the package is the smallest part of it. The evening goes into knowing what you run, getting a fix that exists for it, restarting things in the right order, and being there when that goes wrong. This is that sequence on both sides.

The job On servers you run On Brewless
Knowing what you run A list of machines and versions you keep up to date yourself, including the ones nobody has looked at in a year. The manifest and the release say what the image holds. There are no machines of yours to find.
Getting the fix Wait for your distribution to backport it. Until that package exists there is nothing to install, and you are left with partial mitigations. Build on a newer base image. Brewless pulls it by digest, so the image you tested is the image that runs.
Applying it Install on every machine and restart the service. A reload keeps the old executable running, which is how a patched server stays vulnerable. A deploy. The containers are replaced by new ones; nothing is swapped underneath a running process.
The kernel A package and a reboot on every host, one at a time, with traffic moved away first. An installed kernel is not a running kernel. The provider's machines, the provider's reboot. There is nothing of yours to restart.
The window An evening or a night, picked for when it hurts least, with someone watching it. None. Your containers start and stop all day anyway, so a release is the same action as any other deploy.
When it goes wrong A broken service on a live machine, fixed under time pressure, at the hour you picked for it. Roll back to the previous release, which is still there.

None of that makes an application safer by itself. It moves the work, and the risk of doing it at two in the morning, to someone who does it all day. What it leaves you is the part no provider can do for you.

What runs on Brewless

Brewless deploys your Laravel application to Scaleway Serverless Containers, in your own Scaleway project, with bunny.net in front of it. No web server of yours stands in that path.

FrankenPHP, which is Caddy

The public Brewless base image carries FrankenPHP and PHP, and runs it in classic mode. The web server is part of that image, so it is updated by building a release on a newer base image, not by logging in anywhere.

No host of yours

Scaleway decides which machines your container runs on, starts and stops it, and keeps those machines up to date. There is nothing for you to reboot, and no maintenance window to plan: your containers start and stop all day anyway.

No server in front of it either

A visitor connects to bunny.net, and bunny.net to your container. There is no load balancer or reverse proxy of yours in between, and Brewless itself is never in the path of a request.

A deploy is the patch

Updating the layer below your code is the same action as shipping your code: build and deploy. Brewless pulls the base image by digest, so an image swapped under its name is never built on, and a new base image reaches your application with your next deploy.

What you give up for it

Moving the web server and the kernel to a provider means trusting that provider to patch them. Neither you nor Brewless can see Scaleway's hosts or choose when they are updated. What you get back is the work, the restarts and the planning around them.

Still yours

Serverless moves a layer away; it does not move your application away. What is in the image is yours to keep current.

PHP and its extensions

They come from the Brewless base image. When a new one is published, your next deploy picks it up. Nothing updates itself under a running container.

Your Composer packages

composer audit names the ones with a known flaw. Brewless runs it in its own pipeline, and a deploy is how a fix reaches production.

Everything you still run yourself

A CI runner, a self-hosted database, an old VPS that still answers on a domain you forgot about. Those are servers like any other, and every advisory about a web server or a kernel is about them.

Your own code

No provider patches a missing authorisation check, an open redirect or a file upload that accepts anything. That layer never moves.

Questions

Who patches the web server under a serverless container?
Nobody logs in to patch it: the web server is part of the image you deploy. On Brewless that is FrankenPHP in the public base image, and it is updated by building a new release on a newer base image. The machines underneath, and their kernel, belong to Scaleway.
Does my application run nginx on Brewless?
No. The public base image carries FrankenPHP, which is Caddy, and your containers run that. bunny.net sits in front of them at the edge, and Brewless itself is never in the path of a request, so an advisory about nginx is not about your containers. It is still about any server you run beside your application, such as a CI runner or a self-hosted database.
Does a container protect me from a kernel flaw?
Not by itself. A container shares the kernel of the host it runs on, so a kernel flaw is a flaw on every container of that host at once. Scaleway documents that each Serverless Container also runs in a sandbox of its own, with gVisor or Kata Containers, rather than straight on the host kernel.
Does serverless mean I never have to patch anything?
No. The web server and the kernel move to the provider, but the image stays yours. PHP, its extensions and your Composer packages still need updates, composer audit still has something to say, and your own code is still your own code.
How long does a security update take on Brewless?
As long as a deploy. There is no package to wait for on your side, no service to restart and no machine to reboot: you build a release and it goes live, and the previous release stays there to roll back to.

Read more