Jaco Kroon <[email protected]> writes: > On 2026/02/10 19:11, Kristian Nielsen wrote:
>> My point was that in the backup scenario - assuming the kernel doesn't crash >> during the backup - the snapshot should be consistent without the fsync to >> disk, the write to the kernel buffers is enough. > Two layers of caching/buffers ... filesystem and block layer. > > Snapshots (LVM) happens at the block layer. If filesystem re-orders > writes even an lvm snapshot could be inconsistent as we've observed > w.r.t. backups without the appropriate safety settings. > > As example, ext4 has a delayed block allocator which means that writes > to filesystem could re-order before writes to block devices. Ok, so snapshots actually require the full cost of fsync() to be consistent :-/ That means --sync-binlog=1 and --innodb-flush-log-at-trx-commit=1. I hadn't realised that, but yes, as you explain it that makes sense, there is no reason for the writes to hit the snapshot in the same order. Thanks! - Kristian. _______________________________________________ discuss mailing list -- [email protected] To unsubscribe send an email to [email protected]
