11–20 of 168 posts
Is this ZOL tip or Linux specific? FWIW, the bug does not seem to reproduce on FreeBSD-current (r332158). However, it also does not reproduce on Ubuntu 4.4.0-116-generic running the ZFS stuff from Ubuntu.
I am not sure if the bad patch was ported to the other OpenZFS platforms.
Why is the thread title hyphenated? It made me expect there's a utility called "zfs-bug" causing data loss.
basically the reason I like my filesystems a few decades old and mature, I would not trust zfs or btrfs with anything critical.
basically the reason I like my filesystems a few decades old and mature, I would not trust zfs or btrfs with anything critical.
Are you still using ext2 then? Ext3 was 2001, and ZFS was 2005. JFS2 is only 19 years old so doesn't quite meet the "decades" requirement.
basically the reason I like my filesystems a few decades old and mature, I would not trust zfs or btrfs with anything critical.
Why is the thread title hyphenated? It made me expect there's a utility called "zfs-bug" causing data loss.
The OP appears to speak German natively. In German they are much more hyphen happy than English. It's probably the most common error I see among German speakers typing English.
It's a very hard habit for them to break. :)
We are working on it. We know what patch introduced the regression and 0.7.8 is going out soon to revert it. Until then, users should downgrade to 0.7.6 if they have not already. The Gentoo and EPEL maintainers have pulled the affected releases from the repositories (technically masked on Gentoo). Ubuntu was never affected. The regression makes it so that creating a new file could fail with ENOSPC after which files c…
How should people behave right now?
Will normal usage of production filesystems erase data, or will read/write activity leave the potentially-orphaned files in place?
You've also mentioned snapshots being tricky in the thread. Should people stop creating snapshots in case orphaned files are not included in the snapshots?
--
> It is incredibly difficult to reproduce on systems running coreutils 8.23 or later.
IIUC:
- This is specifically due to the fact that `cp` in 8.23 is optimized (8.22 created files in {0..2000} order, 8.23+ randomized the order (I don't quite understand why))
- The script in https://gist.github.com/trisk/9966159914d9d5cd5772e44885112d... uses `touch` to create files in random order and some people reported this triggered the bug
We are working on it. We know what patch introduced the regression and 0.7.8 is going out soon to revert it. Until then, users should downgrade to 0.7.6 if they have not already. The Gentoo and EPEL maintainers have pulled the affected releases from the repositories (technically masked on Gentoo). Ubuntu was never affected. The regression makes it so that creating a new file could fail with ENOSPC after which files c…
> We will likely have some way to recover the orphaned files (like ext4’s lost+found) and fix the directory sizes in the very near future. How should people behave right now? Will normal usage of production filesystems erase data, or will read/write activity leave the potentially-orphaned files in place? You've also mentioned snapshots being tricky in the thread. Should people stop creating snapshots in case orphaned…
Is this ZOL tip or Linux specific? FWIW, the bug does not seem to reproduce on FreeBSD-current (r332158). However, it also does not reproduce on Ubuntu 4.4.0-116-generic running the ZFS stuff from Ubuntu.
Ubuntu was never affected. The regression started in 0.7.7 and Ubuntu is on 0.7.5. HEAD was affected until earlier today when the patch was reverted. I am not sure if the bad patch was ported to the other OpenZFS platforms.
It also reminds me why my NAS runs Debian..