I don’t know, systems are quite complex these days. With all the RAM caching and write queuing it is common that the beginning of a file transfer is very fast. Later, when the data actually is written to the disks, transfers usually slow down. But they shoudn’t get stuck completly.
Also, check dmesg. Always useful if you suspect something is going wrong.
i was randomly looking in the Hardware and maybe i know why its so slow. the new 12tb is on another Controller (Onboard, not the SAS Card) than the first one. Maybe this is the reason, why the perfomance is so low.
Just a quick question since i don’t see it mentioned anywhere:
What Drives do you use and did you check before buying them if they are cmr or smr drives?
My 8TB Drives with high speed (test 280mib write with 2 drives) are wd80efzx = CMR
My 2 12TB are 2x WD120EFAX = CMR
i checked this on nascompares.
Little Update:
I now plugged on all HDD Drives on my HBA Controller ( 0Gtek 12G Internal PCI-E SAS/SATA HBA Controller Card, Broadcom’s SAS 3008, Compatible for SAS 9300-8I)
Same behavior, the 12tbs writing only with 40mib/s
My 3x 4tb Home writes with 129mib/s
My 2x 8tb runs with 280mib/s. The only special here is, that i have a SSD as metadata vdev.
Iam confused why the 2 12TBs only writes with 40mib/s
Just to be sure: are you aware that both your pools are stripes, not mirrors?
If any of your disks should die you will lose all data on the pool that it’s a member of.
Please post the SMART values of at least the backup disks here, even better: the values of all disks in the system (e.g. by running for disk in $(smartctl --scan | cut -d" " -f1); do echo " === $disk === "; smartctl -x $disk; done in bash).
That doesn’t explain the bad performance of course. If anything, performance should improve for a stripe compared to a mirror.