> On 14 Apr 2023, at 9:38, Evgeny Morozov <postgresql3@xxxxxxxxxxxxxxxxx> wrote: (…) > I don't know whether ZFS zero-fills blocks on disk errors. As I > understood, ZFS should have been able to recover from disk errors (that > were "unrecoverable" at the hardware level) using the data on the other > two disks (which did not report any errors). Thus, PG should not have > seen any corrupted data (if ZFS was working correctly). > https://unix.stackexchange.com/questions/341614/understanding-the-error-reporting-of-zfs-on-linux > seems to confirm this. Am I misunderstanding something? Your problem coincides with a thread at freebsd-current with very similar data corruption after a recent OpenZFS import: blocks of all zeroes, but also missing files. So, perhaps these problems are related? Apparently, there was a recent fix for a data corruption issue with the block_cloning feature enabled, but people are still seeing corruption even when they never enabled that feature. I couldn’t really find the start of the thread in the archives, so this one kind of jumps into the middle of the thread at a relevant-looking point: https://lists.freebsd.org/archives/freebsd-current/2023-April/003446.html Regards, Alban Hertroys -- There is always an exception to always.
