Public bug reported: SRU Justification
Impact: LXD on Ubuntu runs on top of zfs by defaults. Users that make use of shiftfs for efficient id-shifting currently hit a bug where zfs is confused about the amount of space that is used in a dataset. For example, creating a file with 1GB of random data will increase the space used by the dataset by 1GB. When the file is removed via rm the space is not freed for zfs. This leads to zfs running out of space pretty quickly. This bug has been observed, described, and reproduced here https://discuss.linuxcontainers.org/t/trying-out-shiftfs/5155/9 . Stéphane Graber observed related issues. Regression Potential: Limited to shiftfs. This patch has been tested on various backends btrfs, dir, zfs to verify that it doesn't regress other workloads. Shiftfs now also aligns more closely with overlayfs on file deletion. Test Case: sudo snap install lxd sudo snap set lxd shiftfs.enable=true sudo systemctl restart snap.lxd.daemon sudo lxd init # make sure to select zfs as backend sudo lxc launch images:ubuntu/bionic b1 sudo lxc exec b1 -- dd if=/dev/urandom bs=1M count=1000 of=dummy.file sudo zfs list default/containers/b1 # will show +1GB sudo lxc exec b1 -- rm dummy.file sudo zfs list default/containers/b1 # will show +1GB on a non-fixed kernel and -1GB on a fixed kernel Target Kernels: All LTS kernels with shiftfs support. ** Affects: linux (Ubuntu) Importance: Undecided Assignee: Christian Brauner (cbrauner) Status: In Progress ** Changed in: linux (Ubuntu) Assignee: (unassigned) => Christian Brauner (cbrauner) ** Changed in: linux (Ubuntu) Status: New => In Progress -- You received this bug notification because you are a member of Kernel Packages, which is subscribed to linux in Ubuntu. https://bugs.launchpad.net/bugs/1841977 Title: shiftfs: drop entries from cache on unlink Status in linux package in Ubuntu: In Progress Bug description: SRU Justification Impact: LXD on Ubuntu runs on top of zfs by defaults. Users that make use of shiftfs for efficient id-shifting currently hit a bug where zfs is confused about the amount of space that is used in a dataset. For example, creating a file with 1GB of random data will increase the space used by the dataset by 1GB. When the file is removed via rm the space is not freed for zfs. This leads to zfs running out of space pretty quickly. This bug has been observed, described, and reproduced here https://discuss.linuxcontainers.org/t/trying-out-shiftfs/5155/9 . Stéphane Graber observed related issues. Regression Potential: Limited to shiftfs. This patch has been tested on various backends btrfs, dir, zfs to verify that it doesn't regress other workloads. Shiftfs now also aligns more closely with overlayfs on file deletion. Test Case: sudo snap install lxd sudo snap set lxd shiftfs.enable=true sudo systemctl restart snap.lxd.daemon sudo lxd init # make sure to select zfs as backend sudo lxc launch images:ubuntu/bionic b1 sudo lxc exec b1 -- dd if=/dev/urandom bs=1M count=1000 of=dummy.file sudo zfs list default/containers/b1 # will show +1GB sudo lxc exec b1 -- rm dummy.file sudo zfs list default/containers/b1 # will show +1GB on a non-fixed kernel and -1GB on a fixed kernel Target Kernels: All LTS kernels with shiftfs support. To manage notifications about this bug go to: https://bugs.launchpad.net/ubuntu/+source/linux/+bug/1841977/+subscriptions -- Mailing list: https://launchpad.net/~kernel-packages Post to : kernel-packages@lists.launchpad.net Unsubscribe : https://launchpad.net/~kernel-packages More help : https://help.launchpad.net/ListHelp
