Exercise 3: A ZFS Pool That Fails to Import After a Kernel Update — Possible Solution ==================================================================== Likely cause: Per the chapter's own warn-box, "because the ZFS module builds against the currently-installed kernel via DKMS, a kernel update can leave the new kernel without a working ZFS module until DKMS finishes rebuilding it -- zpool import can fail entirely on the new kernel in the meantime." A routine kernel update followed by a reboot into the new kernel is exactly this scenario: the new kernel has booted successfully, but the out-of-tree ZFS kernel module has not yet been rebuilt (or failed to rebuild) against it, so the zfs/zpool commands have no working kernel module to talk to at all -- which explains why the pool fails to import with no disk-level error: the disks and the pool data on them are very likely completely fine, it's the kernel-module layer above them that isn't ready yet. This is a direct, practical consequence of the licensing situation the chapter names -- because ZFS can't be built into the mainline kernel itself due to CDDL/GPL incompatibility, it depends on DKMS rebuilding it separately for every new kernel version, which introduces exactly this kind of timing gap that a mainline-integrated filesystem like Btrfs would not have. Correct next step: Per the chapter's own explicit cross-reference, "grub1-4's own live-edit skill applies directly: boot into the previous kernel from the GRUB menu while the module catches up, rather than assuming ZFS itself is broken." Rebooting and selecting the previous, known-working kernel from the GRUB menu restores a kernel the ZFS DKMS module was already successfully built against, allowing the pool to import normally again while the DKMS rebuild for the new kernel is investigated or allowed to complete, rather than treating this as data loss or pool corruption. WHY THIS WORKS AS AN ANSWER ------------------------------ This identifies the actual mechanism (a DKMS module rebuild gap after a kernel update, not a disk or pool problem) using the chapter's own warn-box, and gives the specific correct recovery step the chapter names (booting the previous kernel via GRUB) rather than a generic "reinstall ZFS" guess.