I also run Debian (almost) everywhere. For those who want a rolling release, Debian testing is also surprisingly stable. One of my laptops has been on testing getting continuous updates for about 7 years now and still runs great.

Even Debian unstable is reasonably stable for me (despite the name), comparable to other rolling-release distros I think. Really not-ready-to-ship stuff is uploaded to 'experimental' instead. The main precaution I take is having 'apt-listbugs' installed. It checks for any new high-severity bugs in the packages you're upgrading. You can take a look at them before deciding whether to upgrade, which occasionally helps avoid installing something broken.

On servers I prefer Debian stable though. I like my mail server running unattended with only security upgrades, and rolling-release distros tend to introduce config-breaking changes exactly when I have the least time to deal with them.

You're describing my setup exactly. Stable for servers, testing for personal machines, with apt-listbugs installed (the install guide recommended this at one point, not sure if it still does). I also ran unstable for about two years and, like you said, it was fine.

Testing is usually only a few weeks behind unstable though, so if your primary motivation is to have relatively recent software and a rolling release so that you never have to stop everything and deal with full system upgrade breakages, then testing is great.

How do you apply kernel update without rebooting the server?

There are a lot of live patching solutions available, paid and free. On my Debian systems, I use KernelCare (like $2.50/server/mo) for a hands-off experience, have only had good experiences over the years.

It's possible that you can do it yourself on Debian (maybe look into https://packages.debian.org/stable/kernel/kpatch), but it was a bit labor intensive, last time I checked, as you have to manually select patches to apply.

We all know OP didn't updated his server for 3 years+

I don't see anything in OPs post that implies that.

Nope, but it's so likely, it may as well be assumed.

1000 days of uptime means the server never rebooted for 3 years, since you need to reboot to apply kernel update / firmware update. ( and plz don't mention ksplice ect .. those are not intended for that )

They absolutely are intended for that. RHEL, Canonical, Oracle, SUSE and KC all disagree with you.

How do you update firmware without rebooting for 3 years? And no it's a stupid idea to use Ksplice for that long and I'm 99.9% sure that OP didn't use Ksplice or similar solution anyway, it's just bad mgmt.

I would argue that’s mostly due to the Linux kernel being stable. Not to take anything away from Debian the hard work of the maintainers and contributed of Debian.

I know this sort of personally aggressive comment is common in some technical circles, but it's not ok on HN. Would you mind reviewing https://news.ycombinator.com/newsguidelines.html and sticking to those rules, and spirit, when posting here? We'd appreciate it.

I second this, Debian is the professionals stable platform. Never does an update leave me F*%@, as I've mentioned in this same thread I'm a Devuan guy (sorry to be repetitive) but I simply cannot get down with systemD - maybe I am just old... Devuan is great, but much much much thanks to the Debian guys for provide a beautiful and solid base. Thanks Debian team you guys are amazing!

You don't have to use systemd on Debian.

"You don't have to use systemd on Debian."

...unless you want a fully functioning desktop machine.

You only have to look at the effort put in by the MX Linux team to keep the shim functioning to realise a non-server Debian system without systemd is a headache for a ordinary user.