ZFS on Hardware RAID

[Gary Palmer]

On Mon, Jan 21, 2019 at 03:17:08PM +0100, Borja Marcos wrote:
> 
> 
> > On 20 Jan 2019, at 16:33, jdelisle <jdelisle at gmail.com> wrote:
> > I can say for certain that with Dell PERC 6/e RAID (SAS 8087ELP based)
> > cards, when the virtual devices are entire disks each in a RAID0
> > configuration, the way the adapter writes data to them is effectively the
> > same as JBOD.  You can take a RAID0 drive from a PERC 6/e controller and
> > use in with an HBA/ "IT Mode" JBOD controller without issue.  I personally
> > migrated 14x 1TB drives off PERC 6/e (all RAID0's) to a SAS 6gbps HBA, and
> > my ZFS VDEV's and pools were totally fine.  You can move back and forth no
> > problem.  I can't say this is the case for ALL RAID controllers, I've only
> > done this with an old PERC 6/e.
> 
> However a single disk RAID0 volume (what manufacturers call a JBOD) is not the same 
> as a disk. With the ???virtual JBOD??? you lose direct CAM access, which means that some
> advanced options won???t be supported.
> 
> For example, if you are using SSDs ZFS knows when to issue a TRIM (or, in general,
> BIO_DELETE) command. Does the ???virtual JBOD??? offer that option? 
> 
> The same could be said of other functionalities available when you talk directly
> to the disks.

This is a pertinent point.  If I remember right, no LSI RAID controller
supports TRIM/UNMAP on SSDs.  Spinning rust is OK in a RAID controller,
but if you have SSDs attached to the LSI RAID controller and plan to use
them in ZFS it is better to get a HBA instead.

I don't know about other RAID controller manufacturers.

Regards,

Gary