Using ReFS integrity checking

chell · Wed Sep 09, 2020 9:45 am

I recently had a database sitting on a starwind volume corrupt after one of the starwind nodes blue screened.
I was thinking of changing this storage to ReFS with integrity checking turned on to prevent the corruption occurring again.
The question is where should I use ReFS and where to enable the integrity checking. There is effectively 3 file systems being used here. All of them are NTFS currently.
1) The local storage that holds the starwind data files
2) The starwind ISCSI drives that is attached as clustered shared volumes. (thick provisioned)
3) HyperV hard disk files that is stored on those clustered shared volumes

Thu Sep 10, 2020 11:24 am

Hi,

Consider redoing everything formatted as NTFS. Sounds like a lot of work to do but NTFS has proven to be more solid at this point.
Take a look at this post for more info https://community.spiceworks.com/topic/ ... d-by-a-san.

chell · Fri Sep 11, 2020 2:05 am

All of the drives are NTFS already. I've had more than one corruption occur to these server. The only difference between these servers and others I manage is that these servers use starwind.

Fri Sep 11, 2020 7:19 am

Did you re-format the StarWind disks as ReFS as well?

chell · Fri Sep 11, 2020 3:42 pm

I haven't done anything yet. Everything is still NTFS and starwind is thick provisioned.

Sun Sep 13, 2020 12:15 am

Could you tell me more about configuration? Is it a hyperconverged (VMs/roles are running on the same box where storage is located) setup? How many physical links do you have (a network diagram will be helpful), what is the underlying storage configuration, are sync and iSCSI connected directly? Is SQL running as a role in your cluster or inside the VM?
If you still want go with ReFS, the local storage that holds the starwind data files can be formatted as ReFS.