Scrub Mechanics

[Mr_N]

Just curious about the mechanics of the zfs scrub process as I'm seeing reading and writing with the first scrub in progress since i added my 2nd vdev to pool.

I have re-copied alot of data to the pool since so all the data isnt just on the original vdev.

So at the start the graphs just showed reads (100-150MB/s) to the original vdev and nothing happening on the new vdev (first hour or so) but since then there has been writing (only around 100kB/s) to the new vdev and continued reading on original.

All the previous scrubs have been completed with single vdev and showed reading for the whole 5-6 hrs...

 

[Ericloewe]

ZFS iterates over all the block pointers. It's normal that old data is first in line, with the newer stuff being towards the end.

 

[Mr_N]

Yeah but why the writing to the new vdev during that 2-3rd hrs prior to the reading of its data towards the end?
Is it occurring on the other vdev also and i just cant see it on the graph due to the reading being of such greater magnitude?

 

[Ericloewe]

Yeah but why the writing to the new vdev during that 2-3rd hrs prior to the reading of its data towards the end?
Is it occurring on the other vdev also and i just cant see it on the graph due to the reading being of such greater magnitude?

Probably unrelated. .system dataset, shares being accessed, ...

 

[Mr_N]

Nope, nothing else was using it during scrub and there were no writes occurring to newer vdev until the scrub started at which point it was continuous...