On 2016-12-31 20:18, Richard Brown wrote:
On 31 December 2016 at 15:42, Carlos E. R.
wrote: Did you do all of the above before trying btrfs check --repair?
Well, I find that procedure too complex. One of the reasons I refuse to use btrfs.
While I understand that, I see the problem as the price of progress
The same way that the correct management of an mdraid or LVM volume group requires more tooling and expertise than a traditional partition, btrfs (which is a filesystem with functionality matching and exceeding in some areas those of LVM and mdraid) has a similar requirement for more tooling and expertise when resolving issues on it
I also go away from LVM... unneeded complexity, difficulty of disaster recovery on my own. Raid does not justify itself for most uses, IMO. A backup strategy is better.
I think if people use btrfs and expect a 'basic' ext4-like experience, then they're going to be sorely disappointed.
Well, I expect a filesystem to "just work".
But on the flipside, the additional features of btrfs enable very broad features with a system-wide impact, such as snapshot and restore, or transactional updates.
I recognize and want the features of btrfs. But I don't trust it, from experience, so I have to say "not for me". I don't actively discourage others from trying, even novices. No. But if they do ask my opinion, then I have to give pros and cons honestly.
And before people point out that XFS is on the way to having some of those features - sure, and when it does, expect the tooling and expertise problem to raise even higher - XFS already has more concerns with it's repair than traditional ext, which is why fsck for xfs does nothing at all and, just like btrfs, you need to use different tools.
XFS has a well designed tool set, and a mail list where the developers kindly help people with disasters. They helped me more than once, and I helped them track a few bugs.
IMO, btrfs should have a single tool that automatically analyzes a btrfs filesystem and automatically decides on the best course of action, perhaps asking the user some simple questions.
That's a good idea, +1 from me, know anyone who could write it?
You ask me? Obviously that's for filesystem devs. A very specialized kind of dev.
You mention, for instance:
mount -o ro,recovery to mount a filesystem with issues
But you need another running system to do that... and Leap doesn't have a rescue image anymore. The OP mentions having to use a Debian live image instead.
A Leap system that fails to boot its btrfs root automatically goes into its rescue system - which has the tooling required
CLI and small. You only have the tools in the initrd. Insufficient. There a bit more in the install image, but barely so.
And we have a Tumbleweed Live Rescue Image
Yes, but not a Leap one.
So this is not a problem and you're straying into territory that makes it sound like you're just searching for things to complain about :)
Sorry if it sounds like that, but you have to expect more complains than congrats on a public place like this ;-)
A lot of the horror stories I've seen regarding btrfs are due to misuse / abuse of btrfs far more than actual issues or bugs,
Well, you can not expect users to be knowledgeable about filesystem nuances, specially being btrfs that advanced and complex.
Sure I can - there is lots of good documentation about it out these days: https://www.suse.com/documentation/sles-12/stor_admin/data/sec_filesystems_m...
SLE... Better on the openSUSE book. Can't search myself this instant, I have to go. New year celebrations etcetera. :-) -- Cheers / Saludos, Carlos E. R. (from 42.2 x86_64 "Malachite" at Telcontar)