On 3/14/2024 2:56 PM, Bob Bishop wrote:
Probably (I don’t know for sure) these drives have a RAM write cache
and really suck at committing from there to NVM.
ZFS doesnt seem to like it, or, possibly something else in addition
going on. I will start to get errors like this
(da4:mrsas0:0:23:0): READ(10). CDB: 28 00 3c 39 f2 60 00 00 08 00
(da4:mrsas0:0:23:0): CAM status: SCSI Status Error
(da4:mrsas0:0:23:0): SCSI status: OK
(da5:mrsas0:0:24:0): READ(10). CDB: 28 00 28 0b b5 38 00 00 f0 00
(da5:mrsas0:0:24:0): CAM status: SCSI Status Error
(da5:mrsas0:0:24:0): SCSI status: OK
when I have a heavy stream of zfs send | zfs recv going on. Not sure if
its possible to work around this or its some other bad interaction, or
just a plain old bug in the firmware of these SSDs
---Mike
