Linux DM-VDO "Virtual Data Optimizer" Preparing To Land In The Upstream Kernel

[Anux]

Write barriers are incredibly important even on the server with ups system.

Is that so? Most file systems describe barriers in their docs like "turn on if you don't have battery backed raid". Why would the option to turn it off even be there if it is of no use?

I would imagine dm-vdo being painfully slow with write barriers on.

[LtdJorge]

Use btrfs then. Safety and zfs in one sentence doesn't go well.

Ah yes, the "there were a few data corruption issues recently with ZFS but I choose to ignore the ones that plagued BTRFS (early on)" take

[oiaohm]

Ah yes, the "there were a few data corruption issues recently with ZFS but I choose to ignore the ones that plagued BTRFS (early on)" take

https://www.jodybruchon.com/2017/03/07/zfs-wont-save-you-fancy-filesystem-fanatics-need-to-get-a-clue-about-bit-rot-and-raid-5/

Turns out a lot of ZFS sales points under deeper research turned out to be snake oil. Yes the ZFS data corruption bugs did end up proving that the theory that lot of ZFS claims was snake oil was in fact true.

One thing btrfs developers have not in fact done is over claim. ZFS developers right back from Sun have in fact over claimed ZFS capabilities.

Like it or not a lot of early on ZFS faults were ignored for a long time and people were not putting correct backup procedures in place because they trusted the ZFS sales points.

The safety and zfs don't go well in one sentence due to how much incorrect information is out there about ZFS level of data protection and this problem starts with the orginal ZFS developers at sun not in fact understanding the bit rot problem and the different forms of write holes.

Yes newer storage media have created new ways to have write holes over time. Data protection on a ongoing up hill battle.