Re: IBM ServeRAID M5110e and other mfi/mrsas users - call for testing / request for information
- Reply: Ed Maste : "Re: IBM ServeRAID M5110e and other mfi/mrsas users - call for testing / request for information"
- In reply to: Johan Hendriks : "Re: IBM ServeRAID M5110e and other mfi/mrsas users - call for testing / request for information"
- Go to: [ bottom of page ] [ top of archives ] [ this month ]
Date: Tue, 26 Nov 2024 19:58:14 UTC
> On 26 Nov 2024, at 13.31, Johan Hendriks <joh.hendriks@gmail.com> wrote: >> >> To test, set the loader tunable hw.mfi.mrsas_enable=1 in >> /boot/loader.conf. Confirm that mrsas is being used, and that the >> system functions properly. >> > I use Dell servers and i set the loader tuneable hw.mfi.mrsasa_enable=1 on these machines. I use ZFS, so no raid functionality off the card itself. This works fine for me. I have not have any issues. I use this setting as mrsas give me way nicer device names, i do remember having boot issues with the mfi default. But that is a long time ago. I have several Dell servers (R640/R740) running 14-stable using mrsas and RAID10 volumes, UFS on FreeBSD, and I have experienced corrupted filesystems on reboot on 3 out of 14 servers so its not consistent, its the same servers that always fail though :) However for our purposes we have set cache policy to write-through instead of write-back (default) and that seems to have solved the issue. There is no real performance penalty at least with our workload. Would be nice to have this fixed though, seems like the cache is not written to disk on reboot but if you force it through it works. Its just like when the backup battery on the controller dies, then a reboot also messes up the disk :) -Søren