Hacker News Reader

Comment by turnsout | Hacker News Reader

I've been running a FreeBSD server (web, email, DNS) for 20 years at this point, so consider this a long-term review.

It just works! Specifically, keeping the system running and up-to-date is very low-maintenance. On the software side, the pkg system has been great—and a welcome change from manually installing ports from source. On the OS side, security updates and upgrading major and minor versions have all been painless.

I haven't run a Linux server long-term, so I can't compare, but my biased perception from the outside is that it seems like Linux changes more rapidly and requires a bit more hands-on management.

View on HN

That heavily depends on distro.

Debian upgrades "just work". Even had colleague accidentally upgrade 2 releases at once (put too new repo on one of the old servers) and it worked without a hitch

Ubuntu? eh, usually ?

CentOS/RHEL/derivatives ? It isn't even supported as a feature.

Both my desktop and NAS are going 10+ years with just dist-upgrade. Also few hundred servers at work where we migrated from CentOS (which was a chore to manage tbh).

To be fair, Debian used to work even for FreeBSD back in the day with Debian GNU/kFreeBSD.

> To be fair, Debian used to work even for FreeBSD back in the day with Debian GNU/kFreeBSD.

As both a Debian and FreeBSD person, I think I was uniquely in a position to like this, and it was awful circa 2012. Some stuff would work, and a lot of stuff didn't, because FreeBSD basic utilities weren't packaged and GNU versions didn't have full functionality: network configuration was difficult, lots of administration was also difficult. You ended up needing a FreeBSD chroot, and then you might as well use FreeBSD/FreeBSD.

That's interesting about Debian and dist-upgrade. I think I had assumed everyone had consolidated on Ubuntu, but every Ubuntu installation I've run has been a short-lived VPS, so I haven't had to think too much about long-term upgrades. But maybe that's the new mindset—just periodically wipe everything out and reinstall your application from a script/container/what-have-you.

That's certainly a way, especially that reinstall takes about as much as full upgrade.

But I use configuration management anyway to keep the machines consistent, and, well, upgrade is less work than reinstall... especially if talking hardware that can take 5+ minutes just to reboot (which is like every actual enterprise server I saw).

I’m fairly sure RHEL has had in place upgrades for a few release now (at least going back to the 6 -> 7 transition, nearly a decade ago).

Reply on news.ycombinator.com