I had no idea we were discussing servers. I know nothing about them at all. But I would not imagine a rolling release would be necessary hardly ever.
I agree with all the points here but the second. I actually find Arch more stable.
Yeah? How so?
Fwiw, here's two little examples relevant to me.
- i7 ssd + hdd 32 GB ram Tower
- Manjaro KDE from December 2017
- Converted MJ to Cleanjaro KDE several months ago.
- In December 2019 i made my primary boot Arch [via Archfi] KDE, with CJ KDE now my secondary boot.
- With MJ/CJ, i would/do routinely lose all network connectivity each time i Resume, after the night's Suspend. In Arch, that never ever happens, not even once so far.
- With MJ/CJ, i would/do routinely experience the Plasma Network Monitor widget freeze each time i Resume, after the night's Suspend. In Arch, that never ever happens, not even once so far.
I am not in any way disparaging the MJ MHWD tool, but fyi so far in Arch i have never had even a glimmer of hassle managing my kernels [LTS, Zen & Linux], & several updates have occurred already.
This goes to performance not reliability, but in Arch my Tower boots spectacularly fast, whereas MJ took much longer.
This goes to usability not reliability, but in Arch i am absolutely loving
For me, this tends to happen during the times when Manjaro devs hold back certain updates due to apparent problems for the same updated packages in Arch, or when certain updates in Arch would require a user's manual intervention.
Manjaro devs might hold back until updated fixed packages are available or until they can work out a way to automate the manual intervention in Manjaro (since a fair number of Manjaro users expect to NOT have to do any manual intervention).
But this leads to accumulation of updates that aren't released, since the withholding of an update might affect other packages which receive updates, if they are dependent on the withheld packages. The more updates get held back, the more potential for issues when the updates are finally released to the Stable Repo.
By that time the update will be a huge download with many packages. With so many possible interactions between the new updates and your current system, the potential for problems is increased. And troubleshooting might be harder because there are so many more problem vectors.
In Arch, that short period of instability is actually not so bad, because you are meant to check the warnings and do your own manual interventions or fixes (just check the forums - most of the time someone else has already posted the issue and solution). So you carry out the interventions or instructions, and it works out.
Or, if there are no fixes as yet, well, since you're forewarned, you could choose not to update your system for 1 or 2 days, at which point the new fixed packages are likely to be out.
If you update every few days or 1 week at most, the updates are smaller and more manageable, and troubleshooting is easier with fewer possible problem vectors.
[Disclaimer - I did not install Arch from scratch, ie the Arch way. But I have been using and maintaining a Arch-repo only [no third party repos] distro since mid 2017, after Apricity OS died. From end 2015 to mid 2017, my install also had a small repo from the Apricity devs with their own icons and theme, plus a few other things, but those are gone now.]
I think you are overlooking the security perspective of it. If you would use frozen mirrors, it would not be fundamentally different from simply updating your system twice a year. That would leave your server very vulnerable against new exploits.
The reason point release distros are used for servers is that security patches are ported to older software releases. And backporting the patches takes a lot more time and effort than just packaging from the upstream. And you often can't just transfer newer packages from more current branch to the frozen mirrors, because they are often built against newer libraries than what the frozen mirror has, leading to breakage.
If we wanted to have a semi rolling frozen branch, it would probably require
- at least 2-3 people working full time on backporting the security patches to the frozen branch
- severely limiting the amount of packages in that branch to a more manageable pool. (have you noticed how little packages there are in centos or suse repos? This is why)
With cutting edge rolling release, majority of the security is handled directly by the upstream projects. This makes it possible to include huge amount of packages in the repos with a relatively small team. But if a distro isn't using the current upstream release, then a lot of that security responsibility necessarily falls on the packagers. And that is more than what a handful of volunteers can handle.
The updates are daily and much smaller so a lot less to go wrong.
In well over two years I only rolled back one package and it was fixed moments later.
Since then I had one other package that was not acting as expected and waited a few hours and the update was already fixed. Usually a work around is posted in the forums within the hour and the updated release right behind.
Arch installs are normally smaller and the OS seems to boot and run faster as it only install what you want and not the stove and kitchen sink.
But how do you cook & wash-up?
That's indeed a pro.
I have a third machine that runs Arch and used to run Manjaro. I use it like once every few months which is also when I update.
I do always check Arch announces for update info beforehand. Never had issues updating on both distros with it.
That is rather strange to me as I solemn have the issues that you described. It is good practice to read the Announcements before applying major updates. Which I didn't do a few time and rendered my system unable to boot into Manjaro.
So far, my Arch-repo distros have benefited from me reading comments in the Manjaro Unstable Announcements, while my Manjaro installs have benefited from me having a heads-up through updating my Arch-repo installs as to what's coming down the pike for Manjaro.
Something I liked out of the box was that that Gnome flavor ships with great settings, plugins, and theme out of the box. It is delightful!
While I'm well aware that Manjaro is based off of Arch, it is a very different Distro with its own goals then Arch is and does.
What some ways that users who are not developers can help to improve Manjaro? The few ways I can think atm are writing more and better documentation, reporting bugs, and bringing new ideas to the table the developers haven't thought of.
I'm sure that there many other ways help.