Btrfs Switch Postponed To Fedora 17

[liam]

Ok. But you can do rollback without using BTRFS. You could try that?

I debated the advantages of ZFS here, and people told me that ZFS is not needed and you can do rollback on Linux without any fancy filesystems.

Using lvm perhaps. I know that there is a yum plugin that takes a snapshot of the system prior to every transaction group and adds the old image path to grub.
I haven't needed to use it, however, so I don't know well it works.
Other than that, I don't know what people could've been speaking of.

[kebabbert]

Ok. But you can do rollback without using BTRFS. You could try that?

I debated the advantages of ZFS here, and people told me that ZFS is not needed and you can do rollback on Linux without any fancy filesystems.

[liam]

Why dont you just rollback if your Linux upgrade breaks? Ive read that you can do that, from people here.

When you use ZFS and Solaris, every upgrade can be placed in its own separate snapshot. In GRUB you choose which snapshot you want to boot from. So if anything breaks, I just boot from the latest working snapshot, and delete the new failing snapshot. Then I am back to the state where I was before the upgrade. It takes just a reboot, and I type one command to destroy the new non working snapshot "# beadm destroy MyNewSnapshot". That is all. There is no need ever to reinstall Solaris using ZFS. Can you not do something like this, in BTRFS? I thought BTRFS also had snapshots?

grub doesnt do btrfs yet, IIRC

[dolio]

Why dont you just rollback if your Linux upgrade breaks? Ive read that you can do that, from people here.

When you use ZFS and Solaris, every upgrade can be placed in its own separate snapshot. In GRUB you choose which snapshot you want to boot from. So if anything breaks, I just boot from the latest working snapshot, and delete the new failing snapshot. Then I am back to the state where I was before the upgrade. It takes just a reboot, and I type one command to destroy the new non working snapshot "# beadm destroy MyNewSnapshot". That is all. There is no need ever to reinstall Solaris using ZFS. Can you not do something like this, in BTRFS? I thought BTRFS also had snapshots?

I can't reliably use any btrfs rollback functionality if the btrfs driver is crashing within 30 seconds of boot (the boot partition isn't btrfs anyway, so I'm not sure how rollbacks work for that portion of the system; I don't think booting from btrfs is even an option at this point). Or at least, I'd have to boot using a working btrfs driver, at which point I've already fixed the problem, or could fix it by changing telling a different kernel to boot (if I had booted from a CD, for example).

I've heard that Fedora is supposed to have package manager support for snapshots in conjunction with btrfs, but I haven't really looked into it yet. And there's certainly no option by default to boot from a separate snapshot (and as I mentioned, the kernel can't even live on a btrfs partition for booting purposes).

Anyhow, I don't really know if Fedora fancies itself toward the casual user or Linux aficionado end of things (I just wanted to try out Gnome for a change, and Gnome 3 looked more appealing than Unity). If it's the former, sending people to a rescue disk after a routine update would be a pretty massive failure. I had enough experience to get myself out of the situation, but my parents (say) would probably have been left with a non-functioning computer.

[kebabbert]

I've been using btrfs on Fedora 15 (for my root partition, which I don't really care if I have to blow away at some point), and it's been okay, but....

They released a kernel update near the end of July. After installing and rebooting, the btrfs (and only btrfs) driver would crash about 30 seconds after booting, causing anything that touched the relevant partitions to hang. Luckily, I was able to boot up and type fast enough to switch grub back to booting an old kernel, but for a bit, I was thinking it was re-install time.

Someone on the Fedora bug tracker came up with a fix pretty speedily, but it still hasn't been pushed out to the updates as far as I can see. I can definitely understand them not wanting their default file system driver putting people into reinstall/rescue disk situations on routine updates.

Why dont you just rollback if your Linux upgrade breaks? Ive read that you can do that, from people here.

When you use ZFS and Solaris, every upgrade can be placed in its own separate snapshot. In GRUB you choose which snapshot you want to boot from. So if anything breaks, I just boot from the latest working snapshot, and delete the new failing snapshot. Then I am back to the state where I was before the upgrade. It takes just a reboot, and I type one command to destroy the new non working snapshot "# beadm destroy MyNewSnapshot". That is all. There is no need ever to reinstall Solaris using ZFS. Can you not do something like this, in BTRFS? I thought BTRFS also had snapshots?

[rewind]

The biggest and most common concern I've noticed about making btrfs default is the lack of a working fsck tool.

That and the fact that file operations in comparison with ext4 are much slower. Beside that, a simple disaster reboot can force you to reinstall.

[BlueJayofEvil]

The biggest and most common concern I've noticed about making btrfs default is the lack of a working fsck tool.

[marcelosoaressouza]

I had a major headache trying btrfs in root partition with debian lenny (6.0.2.1). Dpkg was very slow (due to sync and fsync calls). I upgraded the kernel to 3.0.1 and the problem persists.

[dolio]

Prime Time Readiness

I've been using btrfs on Fedora 15 (for my root partition, which I don't really care if I have to blow away at some point), and it's been okay, but....

They released a kernel update near the end of July. After installing and rebooting, the btrfs (and only btrfs) driver would crash about 30 seconds after booting, causing anything that touched the relevant partitions to hang. Luckily, I was able to boot up and type fast enough to switch grub back to booting an old kernel, but for a bit, I was thinking it was re-install time.

Someone on the Fedora bug tracker came up with a fix pretty speedily, but it still hasn't been pushed out to the updates as far as I can see. I can definitely understand them not wanting their default file system driver putting people into reinstall/rescue disk situations on routine updates.

[GreatEmerald]

I've been using Btrfs for a while now (optional switch for openSUSE 11.4) and so far I have no regrets. Well, aside the fact that I can't set labels on my Btrfs drives.