ZFS scrubbing slows to a crawl (rebooting helps)

BSDsEnjoyer

I've run scrubbing on my ZFS pool several times, but it always slowed down to a crawl at some point and I couldn't understand the reason. It could slow down to a couple megabytes per second or even slower, more or less freezingl for many hours. The pool is rather large (about 12T at the moment).

I had hopes that it would get back to good speeds (hundreds of megabytes per second) but it never did. Only pausing the scrub and rebooting with unpausing it helped. After rebooting and upausing it gets back to good speeds. This is rather inconvenient. I wonder whether this is a general FreeBSD/ZFS problem or more related to GhostBSD.

I scrub when the system is idle, so there is no any load during the whole scrubbing process.

wdt

I wonder if you send a kill -1 or kill -2 to the process, then what happens?
Does it wake up OR quit? I am sure the quit would be graceful, it's not kill -9
Do you know the command to resume the scrub?
Well then, a cron to check every 10m? and do above
I know almost nothing about zfs, there must be commands
It would be a slightly tricky script, but not over complex

BSDsEnjoyer

wdt I wonder if you send a kill -1 or kill -2 to the process, then what happens?

I haven't tried that. I can pause the scrub by zpool scrub -p [poolname]. It is fully responsive, just very slow. Unpausing it and resuming does not make it faster, only rebooting helps. Another alternative is zpool scrub -s [poolname]. This will nullify the scrub. After such a command and starting the scrub again from scratch, the speed returned to normal, but the problem returned at some point during this new scrubbing, so stopping the scrub completely does not take anywhere and with several terabytes scrubbed, that's a big loss of spent time. Rebooting allows to unpause and then scrubbing starts off from where it paused with a good speed. Thankfully, I won't need to scrub for some time, I'd see how scrub will work out when I'd need it again.