Slashdot Mirror

← Back to Stories (view on slashdot.org)

OpenSUSE May Be First Major Distro To Adopt Btrfs By Default

Posted by timothy on from the slippery dept.

An anonymous reader writes "The openSUSE Linux distribution looks like it may be the first major Linux distribution to ship the Btrfs file-system by default. The openSUSE 13.1 release is due out in November and is still using EXT4 by default, but after that the developers are looking at having openSUSE using Btrfs by default on new installations. The Btrfs features to be enabled would be the ones the developers feel are data-safe."

104 comments

  1. It'll be news once they do it by Strawser · · Score: 3, Insightful

    Not really that interesting that they're "considering" it. Linux produces an endless litany of RSNs that never come to fruition. I've basically become numb to predictions about the future of the system. Everyone's been planning to do everything RSN for a decade and a half.

    --
    The louder he talked of his honour, the faster we counted our spoons. -- Ralph Waldo Emerson
    1. Re:It'll be news once they do it by icebike · · Score: 4, Interesting

      On the other hand, OpenSuse, and SuSE before them, have a track record of adopting newer file systems as the default.
      They also demote some filesystems as the default, (while still making them available for the user to set as the default.).
      (I still use reiserfs on some systems, it may not be massively scale-able, but its pretty bullet proof).

      But more to the point, I can't really understand your point about RSNs, since Btrfs is already available in OpenSuse and several other Distros for the last several releases.

      Further, on Opensuse at least, the user can set any of the choices as the default for any new partitions, or as the system default at install time. The available choices include Btrfs, XFS and Reiserfs, and three versions of Ext.

      Its not that something is promised and not delivered. Its more akin to having the default web browser set to Chrome or Firefox.

      There is no broken promises here. Simply a failure to understand that the choice has been there for years.

      --
      Sig Battery depleted. Reverting to safe mode.
    2. Re:It'll be news once they do it by Strawser · · Score: 1

      Fair enough points. It's just that, over the years, we've seen so many predictions of what's going to happen next, and something completely different does. 20?? will be the year of the linux desktop; there will never be a 3.x kernel; Linux will never be an enterprise quality system; RedHat is going to die (back when people were calling Red Hat the 'Microsoft of Linux' a decade or so ago); etc. We once thought Linux would never be found in enterprise data centers, but I remember the first time we got a linux box in one. Then we thought it wouldn't scale. Then we had whole server rooms full of it. Everything was going to be resiser fs once upon a time, but that's far from happened. Hell, when I started, Slackware was the "most powerful" distro and no one would ever supplant it. Things change so fast, and so wildly, news of future maybes aren't really that moving.

      I suppose a positive comment would have been more valuable. Don't mean to be negative right off the bat (and I'm usually not), it's just that "news" that someone is "considering" doing something with Linux isn't really worth reporting on -- unless it's a hell of a slow news day.

      --
      The louder he talked of his honour, the faster we counted our spoons. -- Ralph Waldo Emerson
    3. Re:It'll be news once they do it by Anonymous Coward · · Score: 0

      (I still use reiserfs on some systems, it may not be massively scale-able, but its pretty bullet proof).

      Too bad the wife wasn't; that basically killed its adoption and development.

    4. Re:It'll be news once they do it by Anonymous Coward · · Score: 0

      The only reason ReiserFS was dropped as the default was because he was convicted of murder. There was no technical reason.

    5. Re:It'll be news once they do it by icebike · · Score: 1

      Black humor or not, that's just funny.

      Still, Suse was doing most of the maintenance on that package for many years before her demise.

      --
      Sig Battery depleted. Reverting to safe mode.
    6. Re:It'll be news once they do it by Anonymous Coward · · Score: 0

      Yeah, it was the "adoption" part that really got hurt by the whole thing. It's a shame that people couldn't separate the filesystem from the person, but I guess that's what happens when you stick your name all over something and then do something horrible. Regardless of any technical merits, how many people would want to use StalinFS on Hitlux, for example?

      (Same AC as before. Glad you liked the joke, I couldn't resist it)

    7. Re:It'll be news once they do it by Anonymous Coward · · Score: 0

      Yeah, it was the "adoption" part that really got hurt by the whole thing. It's a shame that people couldn't separate the filesystem from the person, but I guess that's what happens when you stick your name all over something and then do something horrible.

      Oh, come on. It's simply not true that ReiserFS died only because the entire public was so simpleminded that it thought ReiserFS was stained by the murder. It's not like this was an issue known to the world at large; if someone had wanted to fork and rename, everyone affected would've understood and been fine with it. The real problem was that, contrary to popular slashdot opinion, ReiserFS was circling the drain long before Hans Reiser murdered his wife.

      For example, reread what "icebike" wrote. After adopting ReiserFS, SuSE was forced to do a lot of essential maintenance on its own. That was because Hans Reiser didn't care about anyone but the paid customers of Namesys (the company he founded), and he didn't even do a particularly good job of supporting them. The on-disk format changed a lot between major versions without even a nod towards backwards compatibility. Old versions required to keep mounting and using old disks were left to rot with no maintenance. Few if any versions ever shipped with adequate disaster recovery tools. There were always lots of bugs, some of them hilariously awful. (For a while it was fatal to write a file containing an image of a ReiserFS filesystem to a ReiserFS filesystem. RFS code would interpret RFS data structures in the ISO image as being part of the host RFS filesystem, and the funny thing is that this was a consequence of design choices intended to improve reliability.) So on and so forth.

      Reiser also alienated the Linux kernel development community. Go searching on your favorite LKML archive and you can find plenty of Reiser flamewars, often featuring Reiser making paranoid accusations of grand conspiracies against him. (Usually because someone was pushing back against integrating mountains of shitty, badly architected code into the mainline kernel. Or against Reiser's attempts to get RFS promoted as a more mainstream FS even though it was obviously unstable.) These antics burned bridges with potential outside contributors; the only people working on ReiserFS were a few Russians he'd hired at exploitation rates. (Well, I don't know that they were exploitation rates, but I remember Reiser being smug about how cheap he was getting skilled labor, and flaunting his ability to order them around like peons. So... yeah.)

      Basically the deeper you look into all that history, the more you think "holy shit, why did anyone ever pay the slightest bit of attention to this narcissistic sociopath". And the answer is that he was pretty smart, had a couple interesting ideas, and like many sociopaths he was slick enough at self promotion and gladhanding to get a lot of attention. But he wasn't good enough at personally coding those ideas (or at leading others) to turn them into a useful filesystem, and in fact his personality flaws had caused the actual product to become a mess. So it really wasn't too surprising when nobody was all that interested in picking up the pieces (except, IIRC, the Russians, who tried to keep things going in order to keep the money flowing in, but simply didn't have the same connections or charisma).

    8. Re:It'll be news once they do it by Zontar+The+Mindless · · Score: 0

      The only reason ReiserFS was dropped as the default was because he was convicted of murder. There was no technical reason.

      Bzzzzzzzzzzt--wrong.

      SUSE switched its default away from reiserfs in 2006, which was about 2 years before Hans even got charged with anything, much less convicted.

      (Longtime OpenSUSE user here. I switched to ext3 because SUSE works exactly as icebike says in this regard, so when they changed the default, I tried ext3 as well as jfs, and found that reiserfs sucked balls compared to both. For this reason, I've not used reiserfs in years, and it isn't popular anymore because there are many better alternatives now available, and because it's not been actively developed for at least 5 years.)

      --
      Il n'y a pas de Planet B.
    9. Re:It'll be news once they do it by Anonymous Coward · · Score: 0

      Oh, come on. It's simply not true that ReiserFS died only because the entire public was so simpleminded that it thought ReiserFS was stained by the murder

      I never said that was the only reason it died, just that it's a shame that people disassociate from a product due to a dislike of a related person. It happened there, and you can go look up the news stories about the death of Reiser's wife if you need proof of that: people were swearing off of ReiserFS simply on principle, even if it was just a vocal minority.

      The same thing happens any time there's a story about the latest Torvalds rant, and just TRY to talk about Ender's Game without somebody bringing Card's personal policies into it. People like to attach a face to a product, and in some cases it's extremely easy to do that, often because the person likes the attention. When one of those faces does something bad, many of those same people will associate badness to the product, regardless of whether it's true or not.

      I remember a period prior to the murder drama where ReiserFS was the up-and-coming darling of filesystems, with people clamouring to make it default and replace ext. This became less common after the murder, while people talking about his various other character flaws became more common.

      Bugs are easy to fix; damaged reputation, not so much.

    10. Re: It'll be news once they do it by Shirley+Marquez · · Score: 0

      ReiserFS got dropped largely because Reiser dropped the ball on maintenance and bug fixing. All his attention shifted to Reiser v4 (which, to be fair, was an ambitious attempt to create a filesystem technically superior to everything) to the detriment of the v3 installed base.

    11. Re:It'll be news once they do it by Anonymous Coward · · Score: 0

      Bzzzzzzzzzzt--wrong. ReiserFS was the default until he trial...

    12. Re:It'll be news once they do it by Zontar+The+Mindless · · Score: 0

      Hmmm... Wikipedia appears to agree with my memory. What have you got?

      --
      Il n'y a pas de Planet B.
  2. Who uses the defaults? by Anonymous Coward · · Score: 2, Funny

    They should enable all the worst options by default, that way people will learn to learn what they're doing. It's not like installing an OS is something you just do casually without any thought.

    1. Re:Who uses the defaults? by MightyYar · · Score: 1, Funny

      What's the name of your distribution? I'll be avoiding that one!

      --
      W..w..W - Willy Waterloo washes Warren Wiggins who is washing Waldo Woo.
    2. Re:Who uses the defaults? by nine-times · · Score: 3, Insightful

      Desktop/laptop operating systems should be able to be installed casually without any thought.

    3. Re:Who uses the defaults? by Anonymous Coward · · Score: 1

      Well, they tried to get Darwin, but that name was taken.

      So they just call it NietzschOS

    4. Re:Who uses the defaults? by Anonymous Coward · · Score: 0

      Sounds like Gentoo to me...

    5. Re:Who uses the defaults? by Anonymous Coward · · Score: 0

      Desktop/laptop operating systems should be able to be installed casually without any thought.

      Yep, just like lawnmowers. It's wrong for vendors to suck up my precious attention when "Breaking Bad" is on.

      Shut up, kid, your dog needed a haircut anyway.

    6. Re:Who uses the defaults? by Gothmolly · · Score: 2

      By extension, "productivity" tools like Office and gcc should be able to be used casually without any thought.

      Oh wait, that's not true.

      --
      I want to delete my account but Slashdot doesn't allow it.
    7. Re:Who uses the defaults? by nine-times · · Score: 0

      Can't tell if you're trolling, but yes, an Office suite should be able to be used casually with very little thought. You should be able to drop into a word processor, type something up, print it, and send the file to someone else-- all without having to think much about what you're doing. There can be more extensive features that require thought, but the basics should be pretty obvious.

      But gcc? How is gcc lumped in with Office as 'productivity' software? Most people should never need to know what a compiler is.

  3. Wait a minute ... by Anonymous Coward · · Score: 0, Offtopic

    Before I jump on the butter-fs whatcha-jigger, I gotta know: Did the crazy person who wrote it kill anyone?
     
    That was the ONLY thing stopping me from switching file systems last time.

    1. Re:Wait a minute ... by hawkinspeter · · Score: 0

      Nope - that would be ReiserFS. Or at least the crazy people who wrote this haven't been found out yet.

      --
      You're a temporary arrangement of matter sliding towards oblivion in a cold, uncaring universe
    2. Re:Wait a minute ... by Anonymous Coward · · Score: 0

      The answer would be negative for Reiser too at the time his filesystem was adapted by distros, so clearly it's not a useful indicator.

    3. Re:Wait a minute ... by Ash-Fox · · Score: 1

      It was the killer filesystem... Then distros decided to drop it :(

      --
      Change is certain; progress is not obligatory.
  4. Pronunciation question... by Beardydog · · Score: 4, Funny

    Should I be calling it "Butterface"? Because I am calling it "Butterface."

    1. Re:Pronunciation question... by gandhi_2 · · Score: 1

      The man page clearly states that it in fact pronounced "butterface".

      --
      THL phish sticks
    2. Re:Pronunciation question... by swaq · · Score: 3, Informative

      BetterFS and ButterFS are both correct. http://www.linux.com/news/featured-blogs/167-amanda-mcpherson/22449-a-conversation-with-chris-mason-on-btrfs-the-next-generation-file-system-for-linux

    3. Re:Pronunciation question... by Anonymous Coward · · Score: 1

      I pronouce it BitterFS, regardless of what is correct. I find it much more suitable.

    4. Re:Pronunciation question... by Anonymous Coward · · Score: 2, Funny

      BetterFS and ButterFS are both correct
      I pronouce it BitterFS, regardless of what is correct

      Betty Botter bought some butter. "But," said she, "this butter's bitter. If I put it in my batter it will make my batter bitter." So she bought a bit of butter, better than the bitter butter, pit it in her bit of batter, made her bit of batter better. Uh, FS.

      I'll get my hat....

    5. Re:Pronunciation question... by Anonymous Coward · · Score: 0

      I call it "butterfly system". Cute is popular these days and you have to stay with the times!

  5. exciting. by Anonymous Coward · · Score: 5, Interesting

    I've gotten 4 machines running "native zfs for linux" using the stable ppa for ubuntu server 12.04.

    It has been a truly mixed bag. Like a bag full of with crashed machines. At least the data has survived each time.

    I am genuinely excited at the idea of BTRFs becoming production ready.

    1. Re:exciting. by Anonymous Coward · · Score: 0, Troll

      I am genuinely excited at the idea of BTRFs becoming production ready.

      Ha ha ha ha ha ha BTRFS is not production ready. Not now, not in 5 years time.
      Only a fucking idiot would install BTRFS as its default filesystem. Look for when Red Hat stops using XFS and replacing it with BTRFS in its clients hardware. Then you may "safely" say BTRFS is production ready. Until that time stay the fuck away from alfa filesystem.

    2. Re:exciting. by bill_mcgonigle · · Score: 4, Informative

      Like a bag full of with crashed machines

      You probably ran out of memory. No, seriously, don't try it on a machine with less than 3GB of RAM. It's not optimized for that use case yet (version 0.6.2 is current - 1.0 will be 'ready').

      --
      My God, it's Full of Source!
      OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
    3. Re:exciting. by Ash-Fox · · Score: 2

      Why is it not production ready?

      --
      Change is certain; progress is not obligatory.
    4. Re:exciting. by theskipper · · Score: 2

      I had a few crashes even with 8GB until bumping to 16GB. Since then my zfs server runs flawlessly (and fast). So imho 10GB or higher should be considered minimum. And 2x8GB ECC modules aren't that expensive any more so ECC always.

    5. Re:exciting. by Anonymous Coward · · Score: 0

      Out of memory? Seriously? I've run ZFS on several machines with 512MB to 1GB of RAM and they've all worked smoothly. One of them has been on-line for over a year with no problems running with 1GB of RAM. I don't know why people insist ZFS requires a lot of RAM, it runs fine for me with less than a gig.

    6. Re:exciting. by Anonymous Coward · · Score: 0

      I ran into exactly the same problem with 32GB of RAM. I don't know what the problem is, but it's not memory. My machine would hard lock about every 2-3 days while I was trying to use it.

    7. Re:exciting. by WuphonsReach · · Score: 2

      2x8GB of DDR2 ECC is still expensive when compared to DDR3. The price of DDR2, once you want to get up into the 16-64GB per machine range is enough that you should probably consider scrapping the motherboard & CPU and buying something that uses DDR3.

      --
      Wolde you bothe eate your cake, and have your cake?
    8. Re:exciting. by WuphonsReach · · Score: 2

      I am genuinely excited at the idea of BTRFs becoming production ready.

      Don't hold your breath. I've been watching the btrfs development and it's simply not there yet. A good clue for when it will be considered "production ready" would be when RHEL advertises it as something other then a technical preview. And it's still labeled as experimental in Fedora 19 (released July 2013), even after it was slated to become the default in Fedora 16 (which didn't happen).

      So, maybe it makes it in time to be included in RHEL7 as "ready".

      Although Red Hat is already talking about RHEL7 since 2012 of last year, and they'll probably be using one of the Fedora releases as their base. So unless btrfs makes it into FC20 or FC21 as "ready", I think they might miss the RHEL7 release.

      --
      Wolde you bothe eate your cake, and have your cake?
    9. Re:exciting. by Solandri · · Score: 1

      Can't speak for the Linux version of ZFS, but I run ZFS for FreeBSD in a VM (specifically, FreeNAS 8.0->8.2). My understanding is that after Solaris, the FreeBSD version of ZFS is the most mature. I watched the memory profile for the VM. It would start at about 512 MB, then gradually fill up all the RAM I gave it, then "purge" itself and drop down to 512 MB and repeat.

      Upon researching, it looks like the only feature that's really memory-intensive is deduplication. That keeps checksums for every file on disk in memory, so it can quickly detect if a new file you're writing is a dupe. But when I turned it on, it absolutely crushed performance. Writes which had been going at 60-80 MB/sec dropped to about 5 MB/s. I turned off dedupe.

    10. Re:exciting. by Anonymous Coward · · Score: 0

      Oh, believe me - it's not. I was using it a few months ago on my workstation. My workstation was used to build a large product, so lots of concurrent random IO - both reads and writes. The main problem I experienced were occasional stuck filesystem transactions. So basically the system works well, except that when a process tries to access the filesystem it hangs and it lasts for a few minutes. After that everything is normal again. But to be fair to BitterFS I've never lost any data. I see a huge progress here, but I don't think it will be ready for non critical use in one year and it will probably take additional one or two years till it's ready for critical use.

    11. Re:exciting. by theskipper · · Score: 1

      I actually was referring to DDR3 but after checking, I was completely wrong about price. Late last year Kingston was pretty much the only manufacturer that (sporadically) had 2x8GB ECC on the market. So I snagged it around $170. Thought that it would have dropped significantly in price by now; it hasn't budged. But now there are more suppliers.

      http://www.newegg.com/Product/Product.aspx?Item=N82E16820239117

    12. Re:exciting. by Anonymous Coward · · Score: 0

      I've been using ZFS-Fuse on Ubuntu Server since 2008. It's been rock solid on all of my servers. No mixed bag here.

    13. Re:exciting. by Ash-Fox · · Score: 1

      My workstation was used to build a large product, so lots of concurrent random IO - both reads and writes. The main problem I experienced were occasional stuck filesystem transactions. So basically the system works well, except that when a process tries to access the filesystem it hangs and it lasts for a few minutes. After that everything is normal again. But to be fair to BitterFS I've never lost any data. I see a huge progress here, but I don't think it will be ready for non critical use in one year and it will probably take additional one or two years till it's ready for critical use.

      I've been using it on a server (Ubuntu raring) that makes use of Linux Containers on the file system, the Linux Containers themselves are mostly build servers for different distributions (heavy I/O when compiling). I haven't noticed any applications hanging when trying to read/write. I am curious though, how much RAM do you have on your workstation? My server has 24GiB.

      --
      Change is certain; progress is not obligatory.
    14. Re:exciting. by metamatic · · Score: 1

      Eh, I find RedHat way too conservative. RHEL is still shipping Ruby 1.8, for example, and only has 1.9 in beta, even though 1.8 was EOL 6 months ago and 2.0 is stable.

      --
      GCHQ Quantum Insert installed. If only our tongues were made of glass, how much more careful we would be when we speak
  6. DON'T INSTALL OPENSUSE 13.1 by lkcl · · Score: 1, Informative

    there are too many bugs in btrfs for it to be installed in production:
    https://bugzilla.kernel.org/buglist.cgi?component=btrfs

    especially this one, which has yet to be resolved:
    https://bugzilla.kernel.org/show_bug.cgi?id=60860

    which is a major useability issue. yes i made the mistake of installing btrfs on a live production system.

    1. Re:DON'T INSTALL OPENSUSE 13.1 by Anonymous Coward · · Score: 1

      Why on earth would you put something on a production system that isn't even default on the most bleeding edge of systems?

    2. Re:DON'T INSTALL OPENSUSE 13.1 by Bill+Dimm · · Score: 4, Informative

      According to the summary, OpenSUSE 13.1 is not the one that will default to btrfs, so I don't know why you are saying not to install 13.1.

      The openSUSE 13.1 release is due out in November and is still using EXT4 by default, but after that the developers are looking at having openSUSE using Btrfs by default on new installations.

    3. Re:DON'T INSTALL OPENSUSE 13.1 by icebike · · Score: 3, Interesting

      there are too many bugs in btrfs for it to be installed in production:
      https://bugzilla.kernel.org/buglist.cgi?component=btrfs

      Well, hold on a second here...

      Your list shows 196 bugs with only 36 still un-fixed.
      Yet EXT4 shows 214 bugs with still 34 still un-fixed.

      Yet Ext4 seems to by adopted by world plus dog.

      --
      Sig Battery depleted. Reverting to safe mode.
    4. Re:DON'T INSTALL OPENSUSE 13.1 by cmurf · · Score: 1

      Especially a bug that the bug reporter has failed to provide developer quested information in the past 10 days, and was also using an older (in the context of btrfs anything older than 3.11 is old) kernel when the problem occurred? Especially uninteresting in my opinion.

    5. Re:DON'T INSTALL OPENSUSE 13.1 by Anonymous Coward · · Score: 0

      Not all bugs are created equal.

      Ext4 is mainly minor bugs while btrfs still have many showstopping bugs.

      Btrfs has a build in performance flaw in the spec. Instead of fixing the spec, they actually put in code that works around the flaw.

      It will never be production-ready

    6. Re:DON'T INSTALL OPENSUSE 13.1 by TCM · · Score: 1

      Because that's the current generation of armchair admins for you. Either that, or his "live production system" is really just his basement porn server.

      --
      Of course it runs NetBSD. BTC: 1NT7QvbetmANwaMzhpVL6
    7. Re:DON'T INSTALL OPENSUSE 13.1 by EETech1 · · Score: 0

      Out of curiosity...

      Citation Needed!

    8. Re:DON'T INSTALL OPENSUSE 13.1 by Anonymous Coward · · Score: 0

      Ext4 adopted by many people, but only has 214 filed bugs. Btrfs used by relatively few people and has 196 bugs. If all of the people using ext4 were to start using btrfs, I'm sure the number of filed bugs for btrfs would skyrocket.

    9. Re:DON'T INSTALL OPENSUSE 13.1 by basecastula+ · · Score: 0

      I would like to know as well what specific bug you are talking about. I think it would be relevant to the discussion. Another note, wouldn't making it a default allow more bugs to be discovered? Thus helping it along.

    10. Re:DON'T INSTALL OPENSUSE 13.1 by Anonymous Coward · · Score: 0

      Ponder for how long and by how many ext4 has been in use. Then do the same for btrfs. Then think some more.

      Note: I'm personally looking forward to btrfs becoming more mature and ready for widespread use in as many different computing environments as possible. The idea that it's less bug-ridden than ext4 seems incorrect to me, however. But that's fine. It's in active development, while ext4 is very much mature at this stage.

    11. Re:DON'T INSTALL OPENSUSE 13.1 by icebike · · Score: 1

      No, ext4 is not yet in the mature stage.
      Compare the bug count with ext3 or ext2, or Reiser, or xfs. (just change the last 4 characters in the link up-thread).

      3 or 6 un closed bugs is the norm, and most of them trivial. But ext4 has some serious unresolved issues.

      --
      Sig Battery depleted. Reverting to safe mode.
    12. Re:DON'T INSTALL OPENSUSE 13.1 by Anonymous Coward · · Score: 0

      http://lkml.indiana.edu/hypermail/linux/kernel/1309.1/02981.html

  7. READY TO ROLL !! by Anonymous Coward · · Score: 0

    Thus isolated errors will cause a maximum of 30 seconds of filesystem changes to be lost at next mount . . . sounds good to me !! What value could possibly be lost ?? It is just bits !! No value lost !! NTFS can go suck a rotten egg !! BeeTreeArrEffEss ruelz !!

    Yours,
    Man behind the curtain

  8. No surprise by willoughby · · Score: 4, Informative

    I remember when SuSE was one of the only distros, perhaps the only one, which used reiserfs as the default filesystem. No, there's no punchline. This was when you could buy it in a box (including the little chamelon pin) off the shelf at CompUSA. SuSE has always had a fascination with new filesystems.

    1. Re:No surprise by ebno-10db · · Score: 4, Funny

      I remember when SuSE was one of the only distros, perhaps the only one, which used reiserfs as the default filesystem.

      Big mistake - it almost killed SuSE.

    2. Re:No surprise by Anonymous Coward · · Score: 0

      I remember when SuSE was one of the only distros, perhaps the only one, which used reiserfs as the default filesystem. No, there's no punchline. This was when you could buy it in a box (including the little chamelon pin) off the shelf at CompUSA. SuSE has always had a fascination with new filesystems.

      THIS! In fact, came here to mention this but you beat me to it. Though back when I was buying SuSE off the shelf you got a free ball cap with it, not some crappy pin. Anyway, they were always leaning forward on filesystem technology, usually to their own detriment. We used it back in the early 2000's as a test platform for roll-your-own NAS and has some interesting experiences with it. We had several >TB boxes that were quite stable, but on the occasion where we did have one crap out on us, look out. Just wasn't worth the risk. My guess is that with BTR in the state that it's in, this will be a similar exercise for sysadmins.

    3. Re:No surprise by Ash-Fox · · Score: 1

      Big mistake - it almost killed SuSE.

      No, it did not. The only thing it did was cause a temporary issue for PR and even then it wasn't even close to killing the distribution.

      --
      Change is certain; progress is not obligatory.
    4. Re:No surprise by Ash-Fox · · Score: 1

      We had several >TB boxes that were quite stable, but on the occasion where we did have one crap out on us, look out.

      I had more, didn't experience this any more with ReiserFS than I did with EXT3 (and every time it was a RAID controller issue).

      --
      Change is certain; progress is not obligatory.
    5. Re:No surprise by Anonymous Coward · · Score: 0

      More than what? Just curious as I simply stated "greater than a TB". Is that like "infinity times infinity"?

    6. Re:No surprise by ebno-10db · · Score: 1

      Oops, I was thinking of Hans's wife.

    7. Re:No surprise by Ash-Fox · · Score: 2

      More than several >TB boxes.

      --
      Change is certain; progress is not obligatory.
    8. Re:No surprise by Ash-Fox · · Score: 2

      She was hidden in the dancing trees.

      --
      Change is certain; progress is not obligatory.
    9. Re:No surprise by Windwraith · · Score: 1

      Congraturations, you missed the joke.

    10. Re:No surprise by Ash-Fox · · Score: 1

      Congratulations, you missed the fact that I realized that from the other reply and replied with a relevant joke to that post in kind.

      --
      Change is certain; progress is not obligatory.
    11. Re:No surprise by Windwraith · · Score: 1

      Congratulations, you are a fool that can't take criticism, also took you a reply to realize it was a joke? Someone needed to point it to you? Then you missed the joke, no matter how you try to save face.

    12. Re:No surprise by Ash-Fox · · Score: 1

      Look whose talking. :)

      u mad?

      --
      Change is certain; progress is not obligatory.
    13. Re:No surprise by Windwraith · · Score: 1

      Why would I be mad?

    14. Re:No surprise by Ash-Fox · · Score: 1

      Why would I be mad?

      Congratulations, you missed the joke.

      _

      --
      Change is certain; progress is not obligatory.
  9. We're what 5 generations beyond NTFS now?! by Deliveranc3 · · Score: 1

    Is there a comparison somewhere? Reiser, EXT4, at least 3 others I've forgotten. This produces a lot of incompatibility, for how much actual performance?

    What do we need, a fastest one and a fastest with X one?

    1. Re:We're what 5 generations beyond NTFS now?! by inhuman_4 · · Score: 4, Informative

      Phoronix Benchmarks will give you an idea of the perfomance differences. Btrfs is usually middle of the pack, so nothing to write home about. The big deal with btrfs is the new features like COW, snapshots, filesystem compression, etc. If you are looking for more performance btrfs is not going to impress. If you are looking for better RAID perfomance, snapshots, compression, etc. Then btrfs is going to be huge for linux. It is probably the closest linux will get to having a ZFS clone.

    2. Re:We're what 5 generations beyond NTFS now?! by Anonymous Coward · · Score: 0

      Well, for what it's worth, ZFS on Sun(Oracle) hardware is by far the best I've ever seen. Working with PB of drive space on a daily basis of nearly every vendor flavor, ZFS (especially on ZFS appliances) is absolutely stunning.

    3. Re:We're what 5 generations beyond NTFS now?! by Ash-Fox · · Score: 1

      Well, for what it's worth, ZFS on Sun(Oracle) hardware is by far the best I've ever seen. Working with PB of drive space on a daily basis of nearly every vendor flavor, ZFS (especially on ZFS appliances) is absolutely stunning.

      It's a shame the CLI utilities on Solaris for managing ZFS are so obnoxious and offered little automatic integration with other features of Solaris. Compare BRTFS with it's automatic snapshot capability with Linux Containers verses Solaris Zones and ZFS' automatic snapshot support (or lack of automatic functionality).

      --
      Change is certain; progress is not obligatory.
    4. Re:We're what 5 generations beyond NTFS now?! by randallman · · Score: 2

      Yes. I use it in conjunction with LXC and making clones is instant thanks to the BTRFS snapshots.

    5. Re:We're what 5 generations beyond NTFS now?! by jedidiah · · Score: 1

      If you've never seen what the FS will do during a crash, then how can you even claim to say anything about it?

      The phrase "success based on blind luck" comes to mind.

      --
      A Pirate and a Puritan look the same on a balance sheet.
    6. Re:We're what 5 generations beyond NTFS now?! by perbu · · Score: 1

      What incompatibility? Btrfs is fully POSIX compliant.

      And I'd be curious to hear what's so fantastic about NTFS.

    7. Re:We're what 5 generations beyond NTFS now?! by silas_moeckel · · Score: 1

      I question the use case, The hardware was defiantly desktop grade and highly memory constrained. 4GB is tinny even for a desktop and a cheap server has at least 8x that. A single SSD again only in a desktop. There are some significant differences in tuning between filesystems, Ext4 was specifically put in ordered mode not writeback for example, when etx4 own docs say write back is faster and the same as xfs etc. Relatime was in use all over the place vs noatime. ZFS shows no mention of ashift=12 being used and the machine barely has enough ram (it's suggested 2GB minimum the BSD freenas suggests 8gb min).

      --
      No sir I dont like it.
    8. Re:We're what 5 generations beyond NTFS now?! by petteyg359 · · Score: 3, Funny

      I question the use case, The hardware was defiantly desktop grade

      Was the hardware told that it absolutely must stop being desktop grade? I see no other reason for it to express defiance.

    9. Re:We're what 5 generations beyond NTFS now?! by Anonymous Coward · · Score: 0

      The problem with ZFS on Solaris is Oracle. With so many other alternatives, using ZFS on Solaris is just not worth the pain of having to deal with Oracle.

    10. Re:We're what 5 generations beyond NTFS now?! by inhuman_4 · · Score: 1

      You do realize that there are hundreds of variables involved here right? Every FS in that benchmarks has lots of tweaks that could make things faster/slower. Plus the differences in kernel version, differences in benchmarks versions, hardware differences, etc. You can't be all things to all of the FS. For example turning on compression for btrfs produces huge performance improvements, but that isn't used either.

      The test lists all of the gritty details I can think of. They may not be optimal, but it better than handwaving the settings or nothing at all.

      If your interested the phoronix_test_suite is FOSS can really easy to run, so if you have some hardware and time, run some benchmarks with a use case you feel is appropriate and let us know how it goes.

    11. Re:We're what 5 generations beyond NTFS now?! by cbhacking · · Score: 2

      Filesystem-level transparent compression, transparent encryption, extended attributes, alternate data streams, integrity levels, multiple ACLs, at least some level of snapshotting, etc. Plus a bunch of stuff that all decent FSes should have, like journaling (not as good as newer FSes, though), symlinks, hardlinks, support for really large (though not ZFS-scale large) volumes, support for really long file and path names, support for many weird characters (prepend \\?\ to a Windows path to use them, as that bypasses Win32 "correction" of path names; just be aware you might not be able to open the file from a Win32 application anymore), and a few other things.

      For its age, NTFS is very good. It is rather old though; there should be better options now (and there arguably are, at least for many use cases). It's not particularly fast, for example, and while you can use it with POSIX permissions (see the Subsystem for Unix Applications in Windows), it's not really built for it (NT only supports one "Owner" which may be a user or a group; SUA has to tack the other ownership info on as an extended attribute). I believe it also still lacks copy-on-write, which is a shame.

      --
      There's no place I could be, since I've found Serenity...
    12. Re:We're what 5 generations beyond NTFS now?! by Anonymous Coward · · Score: 0

      was defiantly desktop

      I doubt that. It was more likely definitely desktop, however.

      Just in case: If you also tend to write "could of" instead of "could have" or "could've" you need to fix that as well.

    13. Re:We're what 5 generations beyond NTFS now?! by Anonymous Coward · · Score: 0

      Filesystem-level transparent compression

      Pretty sure btrfs has got that.

      extended attributes

      And that.

      alternate data streams

      Why is a security nightmare (not to mention a totally illogical way of storing data - just use a freaking standard directory of files) a desirable feature?

      multiple ACLs

      Linux has that option. You can even choose to use one or more different file systems while taking advantage of it.

      at least some level of snapshotting

      btrfs has that, and if you use LVM, you can do it with ANY filesystem.

      Plus a bunch of stuff that all decent FSes should have, like journaling (not as good as newer FSes, though), symlinks, hardlinks, support for really large (though not ZFS-scale large) volumes, support for really long file and path names

      Lots of Linux file systems have all that.

      support for many weird characters (prepend \\?\ to a Windows path to use them, as that bypasses Win32 "correction" of path names; just be aware you might not be able to open the file from a Win32 application anymore)

      You know, I really prefer to just have it take whatever characters I give it. In Linux file systems, I don't need to do any weird crap like that.

      For its age, NTFS is very good. It is rather old though; there should be better options now (and there arguably are, at least for many use cases). It's not particularly fast, for example, and while you can use it with POSIX permissions (see the Subsystem for Unix Applications in Windows), it's not really built for it (NT only supports one "Owner" which may be a user or a group; SUA has to tack the other ownership info on as an extended attribute). I believe it also still lacks copy-on-write, which is a shame.

      Keywords: "For its age". Pretty sure we're talking about now , though. GP implied that progression of file system features is a Bad Thing, and by his subject line, thinks we should all stick to NTFS. So, the question remains - what's so great about NTFS?

      He also implies that somehow different file systems produce incompatibility, which is utter bullshit. Of course you can't read one filesystem using the driver from another (unless it's ext{2,3,4}), but installing another filesystem driver kernel module is a piece of cake.

    14. Re:We're what 5 generations beyond NTFS now?! by cbhacking · · Score: 1

      I read his post and subject line as saying "it's taken this long to catch up to what MS had in the mid 90s?" and in some ways I think he's right. NTFS is far from the be-all-and-end-all of filesystems, but the post I responded to asked "what's so fantastic about NTFS" so I answered. By the way, did you simply forget to address transparent encryption, or did you not have an answer to it?

      Also, ADS is scarcely a security nightmare (no more than, say, symlinks... in fact, less; you can't do trivial TOCTOU attacks with an ADS) although it would be nice if Microsoft had made it easier to show them on file data (it's possible, just harder than it should be). As for why they're a feature, you really have to ask? The ability to keep multiple named, associated, and separate data streams on a single file system entry isn't of obvious use to you? Take the perks of custom extended attributes (which are harder to read under Windows than ADS are, so much for "security nightmare" FUD) and then allow arbitrary length and the ability to open file descriptors to those streams directly.

      --
      There's no place I could be, since I've found Serenity...
    15. Re:We're what 5 generations beyond NTFS now?! by Anonymous Coward · · Score: 0

      Regarding ADS, you're saying it's better/easier to screw around with multiple data segments than it is to just open a file directly? What are you smoking? Use the utterly logical and sensible approach. If you need multiple discrete pieces of data, use a directory with multiple files.

      Encryption is like ACLs and snapshots. It works on ANY filesystem. Just because one OS has stupid anti-modularization that stuffs it into a file system driver doesn't mean that's the only way to do it. Regarding time frames, loop-aes (just one of several encryption options) was released in 2001. Before you go and say "NTFS was around almost a decade before that", NTFS didn't have EFS until 3.0, with Windows 2000.

      More fun dates:

      A version of e2compr (on-the-fly compresion for ext2) was available in 1995, the same year NTFS added compression support.
      NTFS added quota support in 2000. Linux (and thereby all of its file system drivers) had quota sometime between 1996 and 1999 (I cannot find specific commit data).
      NTFS didn't add real symbolic links until the release of Vista.

      So, tell me again, what's so great about NTFS?

  10. Re:Yet another zfs wanabee by bill_mcgonigle · · Score: 2

    Zfs is better.

    For some use cases, yes. For all use cases, of course not.

    What I'm waiting for is a full BTRFS or ZFS-savvy distro layout. And by that, I mean a filesystem for every package with rollback support built into the package managers. Nexenta and Fedora have taken some baby steps in this direction but they only snapshot the whole system at this point.

    "But we can't have six thousand filesystems on a machine!" Of course you can, it's 2013. The FHS was developed for filesystems that existed two decades ago.

    --
    My God, it's Full of Source!
    OUTSIDE_IP=$(dig +short my.ip @outsideip.net)
  11. Re:Yet another zfs wanabee by Anonymous Coward · · Score: 0

    Are you assuming Android or OSX style packages that embed all their dependencies rather than having inter-package dependencies? The package manager is handling dependencies and upgrading or downgrading an entire system image with a consistent set of packages.

    What does it mean to snapshot and/or rollback a package-specific filesystem? How is this different from using a package manager and package repo to upgrade or downgrade packages?

    Do you expect some use of older snapshots other than for restoration? Some kind of namespacing or versioned overlay to allow different processes to run in different logical system images that have different versions of packages, all on the same system? How would this be different from using a package manager to build a chroot that has a particular set of package versions in it? Is it just the cow/dedupe sharing of common data?

  12. Good luck with that by GameboyRMH · · Score: 1

    I hope it turns out better than my experiment with btrfs in early 2012. I can't wait until it's stable and I can use it safely.

    --
    "When information is power, privacy is freedom" - Jah-Wren Ryel
    1. Re:Good luck with that by Anonymous Coward · · Score: 0

      It's been perfectly stable for me on a home desktop, but I doubt I'm stressing it.

  13. Not Recommended by hackus · · Score: 1

    Btrfs has not finalized its disk format yet.

    Until the designers are sure of the final disk layout, I do not think it is wise to adopt it for production use.

    -Hack

    --
    Got Geometrodynamics? Awe, too hard to figure out? Too bad.
    1. Re:Not Recommended by Anonymous Coward · · Score: 0

      Btrfs has not finalized its disk format yet.

      Well...technically true, but as the Wiki says:

      "The filesystem disk format is no longer unstable, and it's not expected to change unless there are strong reasons to do so. If there is a format change, file systems with a unchanged format will continue to be mountable and usable by newer kernels. "

  14. Awesome by ultrasawblade · · Score: 2

    I've been using it awhile, haven't had any problems. Seems to be faster even if it makes my `ps aux` look scary with all those kernel processes.

    Hmm ... that volume didn't have that much free space just a few minutes ago ...

    1. Re:Awesome by gottabeme · · Score: 1

      Btrfs has some special issues with df and subvolumes.

      --
      "Those who consume the bulk of goods are those who make them. We must never forget this secret of our prosperity."
  15. Re: Pronunciation question... PBFS by nbritton · · Score: 1

    Not to be confused with PBFS, a storage medium consisting of ones and zeros written on peanut butter toast. Even with a redundant array of inexpensive peanut buttered toasts the MTTDL is quite high, on account to them being eaten.