HAST + ZFS + NFS + CARP

[Ben RUBSON]

> On 04 Jul 2016, at 20:06, Jordan Hubbard <jkh at ixsystems.com> wrote:
> 
> 
>> On Jul 1, 2016, at 11:23 AM, Ben RUBSON <ben.rubson at gmail.com> wrote:
>> 
>> Would you say that giving an iSCSI disk to ZFS hides some details of the raw disk to ZFS ?
> 
> Yes, of course.
> 
>> I though that iSCSI would have been a totally "transparent" layer, transferring all ZFS requests to the raw disk, giving back the answers, hiding anything.
> 
> Not really, no.  There are other ways of talking to a disk or SSD device, such as getting S.M.A.R.T. data to see when/if a drive is failing.

Yep but SMART is not used by ZFS itself, according to a dev on #openzfs.
There is however a feature request here : https://github.com/zfsonlinux/zfs/issues/2777

I don't know whether FreeBSD iSCSI target implementation is SMART pass-thru or not (I don't think so, my tests some months ago did not work).
However SMART data of iSCSI disks can easily be checked on the target server itself (so not on the initiator, I agree) using smartmontools.

> Drives also return checksum errors that may be masked by the iSCSI target.

Should be caught by smartmontools running on target, right ?

> Finally, there is SCSI-2 and there is SCSI-3 (where things like persistent reservations are implemented).  None of these things are necessarily implemented by iSCSI.

At least FreeBSD implements persistent reservations :
https://www.freebsd.org/cgi/man.cgi?query=ctl
However I'm not sure ZFS itself uses such feature, which are more relevant for clusters. Does it ?

Ben