ECC error reporting importance for ZFS?

[diversity]

LS,

AMD AM4 socket does seem to support ECC but it does not seem to support ECC error reporting. More on this can be found at:
https://forum.proxmox.com/threads/patch-x570-ryzen-edac-support-into-pve-6-1.63744

If ECC error reporting is not functional. Will that pose reliability issues for ZFS?

Kind regards

 

[LnxBil]

Will that pose reliability issues for ZFS?

No, ZFS does not rely neither on ECC nor on reporting. Every filesystem will be better with ECC, but it'll also work without.

Matt Ahrens (one of the ZFS developers) wrote here:

There's nothing special about ZFS that requires/encourages the use of ECC RAM more so than any other filesystem. If you use UFS, EXT, NTFS, btrfs, etc without ECC RAM, you are just as much at risk as if you used ZFS without ECC RAM.

Actually, ZFS can mitigate this risk to some degree if you enable the unsupported ZFS_DEBUG_MODIFY flag (zfs_flags=0x10). This will checksum the data while at rest in memory, and verify it before writing to disk, thus reducing the window of vulnerability from a memory error.

I would simply say: if you love your data, use ECC RAM. Additionally, use a filesystem that checksums your data, such as ZFS.

 

[diversity]

Thank you for the response.

I interpret that quote from Matt differently. More like:
Only when using ECC, ZFS and other filesystems can properly protect your data.

Well now I'd like to rephrase my questions to:
does ZFS also use the ECC error reporting functionality to do it's data repairs or only ECC?

 

[LnxBil]

I interpret that quote from Matt differently. More like:
Only when using ECC, ZFS and other filesystems can properly protect your data.

That is true, but the response was to the often read requirement that you must have ECC for ZFS while other filesystems do not require it. Of course, automatic error correction is better than not having it - as with every additional protection layer.

Well now I'd like to rephrase my questions to:
does ZFS also use the ECC error reporting functionality to do it's data repairs or only ECC?

ECC (without the reporting) is done in hardware, so the OS will have nothing to correct. Failing memory (non-correctable) on the other hand is different, but is also handled by the operating system itself, as it is the glue between the real hardware and the virtual memory it manages. ZFS as a filesystem does only use memory provided by the operating system, so if it cannot rely on the memory provided by the os, you will have to enable ZFS do this - it's in the same quote I gave: use ZFS_DEBUG_MODIFY to checksum your memory too.