OpenSUSE 10.2 will no longer user ReiserFS as its default filesystem. “We’ve been using ReiserFS as our default installation file system for the last 6-7 years now, and it’s served us well in that time. Unfortunately, there are a number of problems with it, some purely technical, some more related to maintenance. I’ll outline a few of the larger issues and offer my solution as a conclusion.”
In the past SUSE did the maintenance on ReiserFS themselves, but I think we can say that ReiserFS (3 that is) is ‘end-of-life’.
Now the switch is made to the relatively safe ext3. That’s a good decision too.
Maybe ReiserFS get’s a new chance when Reiser4 becomes (finally) available in the vanilla kernel.
Definitely a positive move on SUSE’s part from where I’m sitting.
This is quite strange, perhaps I’m an isolated case, but ReiserFS kept destroying my data while ext3 never failed once. So I’d say ext3 is way beyond relatively safe.
…but ReiserFS kept destroying my data while ext3 never failed once. So I’d say ext3 is way beyond relatively safe.
You hear various stories like this from people about all filesystems – ext3 destroyed my data, Reiser destroyed my data, XFS destroyed my data……. They’re utterly meaningless.
They’re generally made by people who get paranoid about what happens if they have a power failure, and they pull the power cord on their machine to test it and whinge when it doesn’t come back up. Such comments don’t apply.
True. On my Powerbook, I’ve run both Linux and OS X, and the latter will regularly destroy its root FS unrecoverably. With Linux, which had to run ext2 for its root FS, and usually had ReiserFS for the others, I’ve only had one unrecoverable error — a minor one on ext2 that made e2fsck segfault, while the FS was still readable. A simple copy, reinitialize and restore fixed the problem.
So my experience says HFS+ w/journaling < ext2 < ReiserFS 3, and the same for their respective fscks. Most other people would say e2fsck is one of the more robust programs out there, and reiserfsck doesn’t have the same reputation. Also, saying that HFS+ with journaling is as stable as nitroglyserin is clearly contradicted by all those Mac users who seem perfectly happy with it.
My guess is that hardware weirdness will have different consequences for different FSs, and ReiserFS isn’t necessarily the worst. My own experience with it has always been positive.
Yabut…..We live in a snow-prone mountain area and in winter have frequent power-outs. Have never suffered data loss on ext3, Reiser or XFS due to these (or anything else), which are all on our server in different partitions for different types of data – so I don’t see that pulling the cord would do it either.
Not really arguing – just a bit confused by your comment.
We also have had quite a few bad issues with reiserfs and we decided now to use ext3 instead. Since then, we haven’t seen that kind of numbers of failures.
To us, it was not ‘meaningless’ – really reiserfs was not useable in our cases.
Similar reports were discussed some time ago with someone who worked for a large european aviation carrier that run linux for some large scale databases. He asked me what I thought of it — told our damaged data and he said “pheww, we’re not the only ones who experience this”.
Maybe ReiserFS get’s a new chance when Reiser4 becomes (finally) available in the vanilla kernel.
Personally I doubt it. As I see it, SUSE has pretty much NO CHOICE and I doubt they’ll back up hans again…IMHO:
– Multicore machines are there, even a normal server can have or will have soon 2,4,8 cores. And this is a short term prediction, in a few years it’ll be 16, 32…
– Reiser3 has know smp scalability problems. Because multicore machines need it, they need a scalable FS NOW. Making reiser scalable would take time, resources, and it’d make reiser very unstable for some time. Aditionally they use beagle, and they may consider to switch to selinux in the future – they need a good extend attributes implementation
– Reiser3 would be scalable if hans reiser had wasted all this years into making it scalable and adding extend attributes
– Instead of doing that, hans dumped all his resources into reiser4. Reiser 4 will take years to stabilize and be usable in enterprises, regardless of if it’s included in the kernel or not and regardless of the engineers you dump on it.
– To make it worse, Hans stopped maintaining reiser3 and flamewared suse when they tried to add features to reiser3. In his opinion, people should “switch to reiser4 now” instead of wasting time adding features to reiser3. This, BTW, made kernel hackers angry. They were left with a filesystem with know problems (reiser3 behaviour with broken hardware is not reasonable) and without maintainers, and users reporting bugs that were fixed by suse but may not get fixed in the future now that suse isn’t there. Don’t get surprised when people doesn’t have warm feelings when reiser asks to include reiser4 in the main tree.
– In other words, hans reiser has created not one, but TWO filesystems that just can NOT be used by serious distributions, even if they want. Reiser 3 has know serious problems that can’t be solved easily and lacks good implementation of neccesary features. And the answer from hans to those problems is…”use reiser4″.
– Consider the alternatives. ext3 has reasonable smp scalability, a good xattr implementation, great behaviour in the case of hardware errors, and all of this is stable, not experimental. Additionally, the developers behave like real engineers and care about fixing real problems in real systems. Ext3 is not the best filesystem on eart but it’s usable for enterprise and ext4 has a reasonable development plan
– Considering all these points, would you seriously put your enterprise distro in the hands of Hans Reiser?
Edited 2006-10-03 22:16
reiserfs v. 3.6 have principial “bugs ?” that cannot be solved w/o disk format change, so Hans stop to develop 3.6 and create reiser4. Sure, all reiser4 bashers knows this fact. So why to continue spreading lie about “Hans laziness to support and patching reiser 3.6”. It fixed, continued and supported. In reiser4.
… Hans Reiser has nothing to do with it.
Reiserfs v3 is broken and will never be fixed (Either due to show-stopper technical problems or due to bad politics – its doesn’t really matter much, once you think about it).
Reiser v4 is years from being sea-worthy.
End Game.
– Gilboa
Now the switch is made to the relatively safe ext3. That’s a good decision too.
It’s a step backwards to be honest, and a sad day for Linux filesystems, certainly for Linux desktop systems. ext3 is a good and reasonably safe filesystem in a server, especially on partitions where not much is happening, while switching to something like XFS for larger partitions for things like serious file serving.
However, ext3 is truly awful on systems used as fileservers and desktop systems where lots of small, or large, files are being accessed. It also creates a crapload more disk activity than any other filesystem I’ve seen – it’s almost pretty scary hearing the drives really go. Watch that LED flash on start up as you wait, and wait for your desktop to boot. ext3 also likes to eat disk space no end.
Oh, and if you want to run a something like a VMware Server, for your own sake put it on a Reiser, or better yet, an XFS filesystem. ext3 will absolutely kill your performance and your hard drive. Alas, JFS is a good filesystem but doesn’t seem to have much support strangely.
You can almost feel Andrew Morton’s pretty clear disappointment on the Linux mailing lists when people talk about extending an already dated filesystem in ext3 for ext4, where he feels Linux deserves a much more modern filesystem where it can really take a lead. Reiser4 may be it, but that seems bogged down completely in technicalities and politics on both sides.
“Watch that LED flash on start up as you wait, and wait for your desktop to boot.”
Since I moved away from reiser3 to ext3 my laptop starts up 5 seconds faster and generally seems quicker even when dealing with small files. However, my reasoning for moving was because I kept ending up with a corrupted file system with reiser3 – this has not been case with ext3 (touch wood).
Ladies and gents, this showcases the power of open source very well. Don’t like the default filesystem? Just slot in a new one, it’s already there and waiting!
They didn’t like Hans Riser’s politics on LKML and now they’ve decided that ReiserFS needs to go. I guess they didn’t also like Reiser’s personal problems with his missing wife.
The other side of the coin is at:
http://kernelnewbies.org/WhyReiser4IsNotIn
Of which I include this short extract (for the sole reason that it’s funny!):
[i]Linux is the kernel that supports many filesystems from its own sources. Linux even includes (being fair) a lot filesystems people consider worthless. Linux supports filesystems that are rarely used….There are several relevant filesystem choices as well….Many people have chosen these and are happy with them. Does that really sounds like a kernel that would not merge a filesystem for political reasons? If anything, Linux looks more like a filesystem bitch.
Does that really sounds like a kernel that would not merge a filesystem for political reasons? If anything, Linux looks more like a filesystem bitch.
The problem is a lot of these filesystems entered the kernel in a state that reiser4 has surpassed already.
The obvious answer is that the kernel developers realized that they had to limit new code entering the kernel because it became a mess.
Unfortunately we’re talking about reiser4 here and Hans Reiser has a way of starting a flamewar whereever he goes =P
Now I’ve read a number of threads on the lkml and imho Hans isn’t the only one to blame here, you get lots of pseudo-arguments, “because I say so”s, polemic statements etc. on both sides and in the end it *has* become a political decision.
That said I hope they manage to overcome their differences soon. I like the philosophy behind reiser 4 =)
I guess they didn’t also like Reiser’s personal problems with his missing wife
Whoa…did you actually read the link? There’s a good number of *technical* reasons that are not related at all with Hans? Did you read anywhere “we didn’t like Reiser’s personal problems with his missing wife”? No, right? Then stop “interpreting” what other people thinks…
One of the big problems with Reiserfs is that since Hans Reiser decided to rewrite it in the form of Reiser 4 (which will take *years* to stabilize and be usable on enterprise distros, regardless of if it’s included or not or how many engineers/money you throw at it) reiser3 has been left without developers, as the link says.
And this will become a problem for reiserfs users in the future. Who is going to maintain it and adapt it to fe.: future VFS changes?
“””I guess they didn’t also like Reiser’s personal problems with his missing wife.”””
No. Reiserfs v3 was abandoned^Wdeclared to be in maintenance mode a long time ago. That’s reason enough.
The “dead wife” problem mainly casts a shadow upon Reiser4.
Hans’ abrasive nature… Namesys’ money problems… Reiser4’s technical problems… and now the search for Nina.
None of it bodes well for the future of Reiser4.
I suppose I may as well throw in Hans’ faked benchmark fiasco:
http://www.uwsg.iu.edu/hypermail/linux/kernel/0408.3/1164.html
Just click “Next in thread” to watch the story unfold. It’s been over 2 years, and the faked benchmarks are still there on the Namesys site, with no notation regarging his reasons for the systematic exclusion of results that favored ext3.
Edited 2006-10-04 21:25
i have always liked reiserfs(3). havent dabled with 4 yet, seems like a pain. ext3 doesnt seem bad and honestly i prolly wouldnt knotice a real world performance difference. i just hate having to change what im use to.
Not to say that XGL is useless (since it provides a lot of code that can be lifted by AIGLX) but I have a feeling that we will be hearing something similar from Novell/SuSE in a few years regarding XGL. I suspect AIGLX will be picked up by the larger community, by the Xorg Project, and by most distributions while XGL will remain something that SuSE has to always pound on to make it work right. The similarities between XGL and Reiser are just too glaring.
Well, I’ve just +1 your comment, some idiot took a point off for a perfect valid comment.
As for XGL/AIGLX and Reiser/ext3fs; its about Novell rationalising what it spends its time and money on; does it make any sense, as a business to prop up things that quite frankly would require them to solely do all the heavy lifting when they could adopt technology that’ll be maintained by a number of different vendors.
AIGLX seems to be delivering everything that XGL promises but it a must less complex manner, and from what it appears has the backing of Xorg, which makes life easier for all concerned.
With all this off Novells chest, they can concentrate on developing and improving Mono, and the parts of the operating system, at the higher levels, that’ll make their product unique and stand out in the marketplace.
Edited 2006-10-04 08:53
A bit OT, but +1.
As long as there was no AiGLX and/or driver support for GLX_texture_from_pixmap Xgl made sense.
Now that AiGLX is maturing and gaining driver support, the dual layer approach of the Xgl seems less appealing.
However, in the long term, having both Xgl and AiGLX being developed is a -good- thing as OSS is all about innovation and choice.
– Gilboa
Yet more people using the *same* filesystem.
This is not because ext3 is great, but because is the one that *everyone* uses. Many people critizes Linux because we have too many filesystems that are good at some thing, but not one that is good at most of things. With this move, ext3 and ext4 become pretty much The Linux Filesystem, and this is good because now SUSE will help to make ext4 a really good fs that matchs XFS and other next-generation fs.
Not so fast; I’m using XFS (amongst others) on one system I have, and exclusively on another.
As for the “too much choice argument”, I don’t fancy having only one filesystem to choose from any more than I fancy one distro. Or OS. Or printer manufacturer. Or hard drive manufacturer. Or computer manufacturer. Or industry. Or party.
Edited 2006-10-03 21:27
Um, I don’t think when people make the “too much choice” argument, they are advocating ONE choice.
I hope not; however the question is, if you’re going to mandate a certain number of filesystems, where do you stop? As far as I can tell each of ext[234]/ReiserFS/XFS/JFS/OCFS serves a different need – even if ReiserFS’s place in the “market” has been taken by ext3, it was created because at the time, ext3 didn’t exist (or wasn’t usable for most people).
All mine are in XFS, rock solid and very fast. Almost went JFS last time thinking maybe SGI’s fall would end support but decided to stick with it.
…but there are a lot of *very* good technical reasons to move on. Hans’ team didn’t maintain it very well after they started concentrating on Rfs4, and the thing still had bugs in it. Maybe when Reiser4 is considered mature, then I’ll try it out again. Until then, the next time I decide to re-install Gentoo, it’ll be on ext3.
I have been using ReiserFS for many many years, but I too will switch to ext3 soon. The problem is that I have an ATI card, and thanks to the %$%$§ driver support my laptop hangs every time I shutdown the PC because the switch between X and console mode is too much for the driver. That has the effect that about once a week I have to boot up knoppix and completely rebuild the reiserfs tree because the filesystem is unbootable.
Gee, that sucks!
Check out the torrents for the dual-architecture dvd available here:
http://www.mandriva.com/en/download/free
I am out to post this. I have an old Mandrake server still working and have very fond memories of Mandrake for the most part. But at some point, I went to Debian and Suse and never looked back.
I am ready now to reassess my options. I hope you guys don’t disappoint. If this release works well in my testing, I’ll recommend it to clients for purchase.
I use XFS on openSUSE on my root and home partitions and it works just great. I just made sure I disabled write-caching so that I wouldn’t have to worry about data loss in the case of a common X-server crash and power outage.
XFS has good performance and has huge capacity limits that beat all other file systems on Linux (that I know of). According to Wikipedia, XFS has the added feature that deleted files cannot be recovered. You can always backup files though. I can defrag my XFS partitions with a command-line tool plus never have to see the time consuming “fsck” operations after twenty or so bootups.
But I agree that Ext3 is a good least common denominator default as long as other file system options remain available.
another XFS irritation for me is the fact you can increase, but not decrease it’s size. kept me on XFS way too long – I had too much data on the XFS drive to be able to copy it to somewhere else, and reformat the whole drive, so i would have had to do it in two steps, resizing the drive inbetween. as you can’t do that with XFS – game over.
What happens if you back up everything on your drive that was formatted in ReiserFS and reinstall SUSE with a different FS? Can your backed up data still be read and written to?
That depends. If you mean “can I read the backup copy just like i used to read my old data,” then that itself depends on whether you copied everything to another filesystem on another disk or used something like afio or tar to back it up; if you did the former, then yes, you can still read it; if you tarred it then you need to untar it onto a new filesystem to read it. (A tar file, originally named for a Tape ARchive, is a file which is an aggregate of whichever files you put into it; this makes it easier to back up onto tapes (obviously!) but also can be useful for copying files from one place to another.)
However, if you mean, for example, “can I back up data on a reiserfs filesystem, reformat it as xfs, then restore the data and read it as normal,” then the answer is a resounding “yes”.
We’ve used JFS for a couple of years now on a large revision control database….and it’s been great, saved my neck a couple of times with recoveries from screwy HDD’s.
SO why did we choose JFS of all things ? We ran extensive tests with IOZone and found that ext3 was slowing the server down too much….there’s certainly a big speed increase with JFS, and so far it’s been extremely reliable (ie. no problems at all) handling our rcs system. Maybe having a heritage from IBM gives it an advantage.
Well, my experience in last 5+ years has been that reiser has never failed me even with a lot of power outages.
On the other hand, I tried ext3 a few times and ended up with corrupted and trashed system after power outage every time.
So I’ve given up on ext3 completely and will never touch it again.
And yes, ext3 is dog slow and it’s anoying how it makes the hard drive trash all the time.
But recently I installed a new small server and decided to use XFS and wow, I was amazed how smooth,quiet, and fast it was.
I’m completely sold on XFS now and will use it on everything, even will convert all my reiser partitions to XFS.
The whole MS IE crap, people bitching that IE came with windows. So what, if you don’t like IE don’t use it, get up & get another browser.