This is going to be difficult for me to make a case for on the technical side as my audience is much more technically knowledgeable than I am. So when I talk about any of the technical aspects of this don’t beat me up to bad if I get something wrong.
f2fs is to new to the Linux universe and that combined with OpenMandriva’s um, lets say limited resources is a bad combination. anything new can and will have problems and I think it would be better to for us to wait until f2fs is more proven to make it our default. BUT: We should (IMO) offer it and make it well know that we offer it with a note in Release Notes about any potential issues we are aware of.
Any performance issues aren’t great enough to justify for average user. Many of you will have read articles and bench tests on this. From my reading while there are measurable performance advantages they aren’t such that most users even a lot of power uses would notice. Nothing profound enough to warrant taking a risk on over more well proven fs like ext4. Note: It’s worth noting if you read the articles and the comments there are a number of people bringing up disadvantages to f2fs on a technical level that I haven’t mentioned here because frankly that lack of knowledge thing again.
Grub2 support is weak and there are issues. This one I’m sure of from my own testing. One thing I ran into was if one multi-boots sometimes OM Lx f2fs systems grub2/os-prober will recognize other Linux systems and sometimes it just won’t. We’ve had multiple user complaints from other users with boot issues while using f2fs that don’t seem to occur with ext4.
f2fs is a headache if one is Multi-booting with other Linux systems. With it os-prober does not consistently recognize other Linux systems and on other Linux systems os-prober won’t see any other systems if even one of them is f2fs. In other words os-prober won’t probe if it sees even one f2fs system. Since I do multi-boot myself I can verify that overall this is a pain in the rear end and I quit trying to deal with it. It’s that much trouble.
Is making a file system designed for NAND/flash devices a default really a good idea when we know we have many users that don’t have SSD’s? We’re seeing a need for i586 or i686 ISO’s so that says to me we have to many users with hardware that may have problems with f2fs. I certainly have read technically knowledgeable people making the case that f2fs is less than ideal for mechanical drives. I myself don’t know if this is fact but it is concerning. Spent a little bit of time before writing this and couldn’t find any information that makes me comfortable with this going on our users mechanical drives until we know more.
This is kind of repeating something said earlier but in helping other users I and others have run into a number of times where using f2fs causes or seemed to us to cause users problems. That’s mainly where this is coming from - this is causing users problems.
Summation: I believe we should at the very least discuss this further hopefully at next TC-meeting. I also believe that keeping ext4 as default file system is our better and arguably the more responsible choice.
What I’m not saying: Not saying f2fs is bad. Not saying anyone did anything bad or wrong. None of that applies on this issue to me.
It was a good idea to try this now we can make a well informed decision.
Incorrect, I’m right on time for a solid decision for Lx 3.02 RC and Final release. We’ve had a year of testing, me included, and there have been issues as reported in first comment above. It seems clear to me that we should offer f2fs for those with NAND/Flash drives but that it should not be default for everyone. We have many users with mechanical drives probably a majority. To many users report issues around grub2 and booting.
I shall see what I can do with bug reports. Edit: Changed my mind, I ain’t doing bug reports on f2fs. Why bother.
IMO the “problem” is that there isn’t a dialog window in calamares at time of choosing file system when users simply go with automatic install (and partitioning and everything).
If there was, so to say, users with SDD would choose f2fs and the others would choose ext4 or what they like best.
Please note: here it’s not complaining about a missing calamares feature, it’s only to explain the situation.
Default settings in general should fit for everyone - or at least for the majority and just a small percentage be out because of strange devices or configuations or such needing of specific changes.
So simple users, who maybe don’t even know what SDD is, maybe not much tech experienced to perform manual partitioning and the likes, those trustful or those “à la windows”: Next > Next > Next > Next > Next > Finish, will end with a file sistem which may cause them some issues and they will shout that OM doesn’t work. Not even knowing where their issues come from
The point here is that what QA are asking for is that f2fst not be the default
filesystem at install not to remove support for it. The post cited at Phoronix
says it’s included and available in the distro it does not bind us to setting
it as the default so it carries no weight in this debate.
Ben has already said that there are problems with it both from the point of
view of spurious error messages at boot but also unreliable multi-booting.
As far as I can see there is no real argument for not having ext4 as the
default filesystem at install should the general consensus think that it is
desirable. The post on Phoronix does not stop us from doing this.
There is a big disadvantage to the current f2fs (which may or may not be a
bug_). If you are a QA person or indeed a developer f2fs is a big timewaster
as if you have a machine lockup and you have to reboot using the power button
you will find that fsck is always run. I have always assumed that since it is
run by systemd that it is essential. If you have a large disk this can take
quite a long time. My experience is that with ext4 (on an ssd disk) this
almost never happens so when crashes occur you can be back up and running very
quickly. If this is not a bug or setup issue then this doesn’t say much for
f2fs’s journaling abilities
Developers who are the most likely people who will want it can f2fs if they so
desire. The fact that it is present is the most important thing.
Why is there such resistance to doing something that makes life easier for
users, testers and developers and in consequence encouraging adoption of the
distro by giving trouble free operation, no error messages reliable multi-
booting and no hanging around waiting for fsck.
If others want to suffer the disadvantages for the benefit of a few seconds
shaved off their boot time and possibly marginally longer disk life then they
can choose to have it.
Best,
Colin
