On 2020-07-01 04:07, Nico Kadel-Garcia wrote:
On Sun, Jun 28, 2020 at 6:21 AM Antti <antti.aspinen@xxxxxxxxx> wrote:
Hello,
I'm in total opposition to this proposal as a long-time Fedora user. The btrfs is unstable and not ready for production. Most of what I'm about to write is admittedly anecdotal but it's the only file system in Linux which has actively and regularly caused me to lose data on desktops, laptops, servers and even on mobile phones when I haven't taken precautions and done regular backups. Something I don't have to actively do when using ext4 in my workstations and notebooks.
We had similar excitement back when reiserfs was popular. My limited
play with btrfs reminds me of reiserfs: feature-filled but fragile and
unsuitable for "/" partitions.
My experience with reiserfs has been very different.
It was a wonderful filesystem, with journaling when nobody had it (ext3 did not
exist, we only had ext2).
It was able to raise the disk capacity of the disk (thanks to tail-packing)
and the performance was great (I was able to immediately notice if the filesystem
was ext2 or reiserfs, as soon as you deleted a big iso: reiserfs was immediate,
thanks to extents vs blocks, I think).
I've never had filesystem corruption on reiserfs even on very hard workloads.
A specific episode remains in my mind: I had rsync hardlink based backups
on a server with software RAID disks. One day I decided to delete about
one year of backups, at four per day it was around 1000 directories, each
of them with 100,000 files (heavily cross hardlinked, of course). To try
to get good parallelism out of the RAID and the elevator, I submitted 1000
rm commands in background and then realized that I was deleting
100,000,000 files with an enormous lock and refcount stress testing on the fs.
After a few hours, operation completed, no issue.
The same production server was switched years later to ext4, since continuing
with a mostly world-forgotten reiserfs option had no point. After a few days
with ext4... fs corruption and data loss, simply because of an online expansion
that was nothing special on reiserfs. Turns out ext4 kernel/tools combination
was bugged.
I've been able to find my comments on bugzilla (year 2012).
https://bugzilla.redhat.com/show_bug.cgi?id=852833#c25
including this:
(BTW, ...spent the weekend restoring a couple terabytes from backups...)
(As a tribute to justice in the world, I want to say, so that search engines index it,
that this system has been in production since 2005 on reiserfs and abused and resized
without any similar issues; it was rebuild recently on new hardware and switched to ext4
and now I really risked to lose the data, as the remote versioned backup is also (resized) ext4,
and the remote-remote backup is also (resized) ext4. I'm not complaining,
but reiserfs really has a much worse reputation than actually deserved).
--
Roberto Ragusa mail at robertoragusa.it
_______________________________________________
devel mailing list -- devel@xxxxxxxxxxxxxxxxxxxxxxx
To unsubscribe send an email to devel-leave@xxxxxxxxxxxxxxxxxxxxxxx
Fedora Code of Conduct: https://docs.fedoraproject.org/en-US/project/code-of-conduct/
List Guidelines: https://fedoraproject.org/wiki/Mailing_list_guidelines
List Archives: https://lists.fedoraproject.org/archives/list/devel@xxxxxxxxxxxxxxxxxxxxxxx
