Announcement

Collapse
No announcement yet.

Google Admin Encourages Trying Btrfs, Not ZFS On Linux

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Google Admin Encourages Trying Btrfs, Not ZFS On Linux

    Phoronix: Google Admin Encourages Trying Btrfs, Not ZFS On Linux

    Last year at LinuxCon a Google administrator was talking up Btrfs and encouraging attendees to try it. That Google admin, Marc Merlin, traveled down to New Zealand last week for LCA2015 to further promote the Btrfs file-system...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    "lukewarm message" In The PDF

    Originally posted by phoronix View Post
    Phoronix: Google Admin Encourages Trying Btrfs, Not ZFS On Linux

    Last year at LinuxCon a Google administrator was talking up Btrfs and encouraging attendees to try it. That Google admin, Marc Merlin, traveled down to New Zealand last week for LCA2015 to further promote the Btrfs file-system...

    http://www.phoronix.com/scan.php?pag...LCA-2015-Btrfs
    I started reading through the PDF. It seemed a reasonable argument until page 11 (at least it was page 11 in my PDF reader) where the stable/safe questions are asked. Then it seemed like the author had to qualify their comments "just a bit". The wording "mostly stable" makes me wonder why I would run Btrfs in "production", or was that comment only in reference to "snapshots, raid 0, raid 1"? It's just difficult to tell from the PDF. Then there were the comments to avoid certain stable kernels; "stay one stable kernel behind" is the advice. In "production" I might actually be further behind than that; some application owners can be "extremely conservative" about the servers hosting their stuff.

    I could go on, but it comes down to this: I just don't get a "warm & fuzz" feeling with regards to Btrfs at this time.

    FTR (for the record):
    I use ZFS on Linux on my storage servers. I typically have 1 ZFS volume per server (works for me in my situation) for "bulk storage"; the OS partitions are "smallish" so ext4 is quite adequate (I could raid1 if I wanted, but never found a real need for it). No problems with ZFS, but kernel upgrades in Gentoo with ZFS present have to be carefully planned to avoid having inaccessible ZFS storage volumes and other "strangeness" occur; I have a process that works everytime so far across multiple servers and 3 kernel upgrades ("gentoo-sources", not "vanilla").

    Comment


    • #3
      Interestingly he didn't find anything actually wrong with ZFS

      Hmm...

      Checking it out, he admitted openly that ZFS is just better and more stable, he just had issues with the license, and then he also said the license incompatibility is not a problem for end users, only for distributions.

      About the memory usage ... just buy more memory .

      Comment


      • #4
        Google...

        Sad that I had a hard time finding google-sponsored commits in the btrfs source tree.
        Kudos there go to mainly Suse, Fujitsu, RedHat, Facebook and Oracle.

        Comment


        • #5
          I'm using Btrfs on my desktop for quite some time now without issues. "Some time" being around 2 years and gone through a convertion from raid 0 to raid5 and then replacing a failing HDD on the raid. everything keeps working as expected.

          Comment


          • #6
            Linux is fast moving and changing, that is one of its major strength. But it has been getting out of hand lately.
            A lot of new code is being pushed that is not tried and tested for years, or a few smart people
            like the new tech and push it hard. Systemd and btrfs are prime examples. I am not against either of them,
            but they are being pushed way to hard into production when they are not ready, in my opinion.

            My experience with ZFS has been amazing on freebsd, on linux it was not stable for me, but after a few months of
            work it is looking really good. In the same amount of time btrfs, has improved but its still rough in a lot of places.
            I am not against using it, but for stability RIGHT NOW, ZFS is the way to go.

            Comment


            • #7
              I'm having a bit of D?j? vu (okay, I'm unsure on the spelling).

              Comment


              • #8
                Originally posted by NotMine999 View Post
                I started reading through the PDF. It seemed a reasonable argument until page 11 (at least it was page 11 in my PDF reader) where the stable/safe questions are asked. Then it seemed like the author had to qualify their comments "just a bit". The wording "mostly stable" makes me wonder why I would run Btrfs in "production", or was that comment only in reference to "snapshots, raid 0, raid 1"? It's just difficult to tell from the PDF. Then there were the comments to avoid certain stable kernels; "stay one stable kernel behind" is the advice. In "production" I might actually be further behind than that; some application owners can be "extremely conservative" about the servers hosting their stuff.

                I could go on, but it comes down to this: I just don't get a "warm & fuzz" feeling with regards to Btrfs at this time.
                Watch the video, it clears things up a bit. In the video he states that 2015 is the year you should start looking at BTRFS and see if it fits your needs. Snapshots, raid 0 and raid 1 should all be working fine. Raid 5 and 6 are still being worked on. "Avoid certain stable kernels" applies to 3.15 and 3.16. 3.14 was fine, then all of a sudden some btrfs bugs got their way into 3.15.x and 3.16.x . 3.17+ is working fine, or so I hear-- this post is being written from a btrfs root on 3.17.8 on Fedora and its working just fine. "Stay on stable kernel behind" is probably just in case a bugfix ends up breaking something but I'm not positive on that.

                The general advice I've seen for btrfs is "run 3.14 or 3.17+", now that could be off or wrong-- but its what I see/hear the most personally.

                As with all things you need to look at it, try it, test it, and see if it fits your quality standards. If so, great, enjoy the safety net btrfs can give you with checksumming and the likes. If it doesn't, okay, great, keep using LVM2+ext4 / ZFS / XFS / whatever-youre-using and look again in a year or two.

                Personal anecdotal note: I've been using BTRFS for about 2yrs now and couldn't be happier. I really love the compression option on my small 120GB SSD
                All opinions are my own not those of my employer if you know who they are.

                Comment


                • #9
                  ZoL Is Tried And Proven For Me

                  ZFS On Linux (ZoL) has been working great for me--serving up several production databases and file systems for 2+ years now.

                  I've never experiences data loss, but early on, there were issues with lockups. Most recently, another nasty lockup https://github.com/zfsonlinux/zfs/issues/2523 was fixed, and ZoL has been running smooth since.

                  There's a lot of mis-information out there still regarding ZFS for Linux like...ZFS relying on FUSE or ZFS requiring tons of memory to run. ZoL is a kernel module with no FUSE layer, and I've run ZoL on machines with several gigabytes to VMs with 1GB or less memory.

                  From a technical perspective of current features, performance, and reliability, I see no reason to chose Btrfs over ZoL . From a license perspective, I don't have an issue installing and loading a CDDL module from source code into my kernels...I've been installing graphic driver binary blobs for years.

                  Comment


                  • #10
                    Originally posted by mgmartin View Post
                    There's a lot of mis-information out there still regarding ZFS for Linux like...ZFS relying on FUSE or ZFS requiring tons of memory to run. ZoL is a kernel module with no FUSE layer, and I've run ZoL on machines with several gigabytes to VMs with 1GB or less memory.
                    To be fair, memory management on ZoL is pretty broken, but there are some patches that look like they may land in the next few months to sort this out considerably. Source: been running ZoL for some years now.

                    Comment

                    Working...
                    X