ZFS srub: last 10% slowing down ...

[psechser]

Hi,

every week I automatically start my scrub on my ZFS pool. I got two vdevs with 4 * 2 TB in RAIDZ1 each.

I experience the first 90% of the scrub run through at about 350 MB/s. But at some 90% and for the last 10% the srub process dramatically slows down to an average of some 250MB/s, which means the last 10% are performing at even less speed (taking some 3 - 4 hours compared to the 90% taking some 7 - 8 hours).

Any idea why this is the case?

thx
Peter

[apollo567]

No idea, but ur scrub time is within the time I had for my pools also.
To be honest, not a real topic to waste too much time on it....

[psechser]

... well, if ZFS is doing something it isn't supposed to do (not working properly as not showing identical behavior throughout the operation) or my disks showing strong behaviors just to mention two things, which I would like to know about, then it would be worthwhile to look into it, wouldn't you agree? It also has something to do with: can I trust ZFS?

[apollo567]

Well, from my perspective yes, simply as there is nothing available which is more advanced.
If you feel worried about ur hardware, check for smart infos from ur disks...

[substr]

Blocksize and write order. If the pool data goes from large block bulk files, to small block zvol that have random write I/O patterns, that will show up in scrub speeds.

If the scrub time is a problem, you would probably be just as well doing it once a month as once a week. Only if you are picking up errors would you need more frequent scrubbing, but you'd also want to be eliminate the source of the errors.

[psechser]

... finally one of my disks said Goodbye and I had to replace it. And my scrub now works nicely ...