I want to get some SAS drives for their added error handling capabilities. My current level of understanding suggests that in order to run the drives with these additional error handling capabilities over the SAS channel means using them in 520/528 byte modes. How do I go about this?
Also, how does ZFS handle this with the whole ashift thing?
There’s a patch for Linux to handle it at a low level but… it doesn’t work that way.
Zfs has its one mechanism that does the same thing so you’d have to write your own patch to make it work.
Oracle database in raw disk mode can use the sector size to write its own error detection codes via the kernel but that’s a narrow use case.
Some raid hba will use more robust error detection and correction routines when used with the larger sector size. This is the main benefit,… hardware raid. In this mode it is faster and less overhead than zfs but still not as robust as zfs.
In short it’s not going to work the way you expect unless you diy it or use it with the appropriate hardware raid controller.
Sorry Wendell, that was way over my head. Are we talking about the same thing?
I’m referring to the T10 DIF/PI and T10 DIX. Reading Seagate articles, it seems that SAS drives come pre formatted with 520byte sectors to make use of the GRD CRC as per this article:
From what I’ve read elsewhere, ZFS isn’t in the same layer and apparently doesn’t even get exposed to the fact that the drive is 520bytes. All it sees is 512byte chunks. Refer to response by Richard Elling:
Wendell is saying if you’re using RAID-z, it will handle all causes of disk errors and nothing else is required. The ESF article you posted indicates that as well, only saying ZFS has CPU and disk I/O overhead that T-10 does not, in case that is critical to your usage. Otherwise, ZFS is better protection than T-10, and does not require special hardware.