ZFS Relationship of slow scrubbing/resilvering and the recordsize/volblock size

B

belon_cfy

Previously my storage was set to 32k recordsize and the scrubbing took about 2-3 days to complete.

Now my server is using 128k recordsize and the average scrubbing takes less than 1 day.

Most of my data on the storage are VMDK images.

Seems the smaller recordsize or volblock will impact the duration of resilvering/scrubbing. If my assumption is right, increasing the recordsize from 32k to 128k will speed up the scrubbing/resilvering up to 4 times faster. Should I increase the recordsize to 128k for fast scrubbing and resilvering?
 
Bigger block sizes will optimize long sequential operations by the cost of short random ones.
 
mav@ said:
Bigger block sizes will optimize long sequential operations by the cost of short random ones.
Click to expand...
That's the reason why scrubbing on small volblock size and recordsize are much slower due to huge random read on storage. I can't imaging how long will it need to scrub a storage with default volblock size which is 8k with the same amount of data, probably 7 days or more .
 
Just migrated all the data to the 128kb block size volume, scrubbing took around 11h37m to complete. Previously it took around 48h up to 81h
 
You must log in or register to reply here.
Back
Top