ZFS corruption related to snapshots post-2.0.x upgrade
a year ago
- #Linux
- #DataCorruption
- #ZFS
- System information includes Debian Buster, Linux Kernel 5.10.0-0.bpo.5-amd64, ZFS Version 2.0.3-1~bpo10+1, and SPL Version 2.0.3-1~bpo10+1.
- Issue observed: Data corruption errors in ZFS send/receive-based backups after enabling crypto, with permanent errors detected in files.
- Error details include a PANIC at arc.c:3790:arc_buf_destroy() during scrubs, leading to system hangs.
- Temporary solution involves rebooting into single-user mode and running scrubs, sometimes requiring multiple attempts.
- Unique aspects of the affected machine: it's a laptop using ZFS crypto (unlike other machines using LUKS).
- Hardware issues ruled out via memory tests and Dell diagnostics.
- Related bugs mentioned: #11688, #11443, and #10697, with some parallels but not exact matches to the described issue.