[zfs-discuss] Replacing drive from JBOD pool

Gordan Bobic gordan.bobic at gmail.com
Tue Dec 23 11:47:06 EST 2014


On Tue, Dec 23, 2014 at 4:39 PM, Michael Kjörling <michael at kjorling.se>
wrote:

> On 23 Dec 2014 15:45 +0000, from gordan.bobic at gmail.com (Gordan Bobic):
> > That, plus lack of support from standard tools like hdparm  or smartctl,
> or
> > requiring use of proprietary tools, is one of the reasons why I prefer to
> > use simple SATA cards with port multipliers. On the whole I find it makes
> > for a much cleaner and more reliable solution.
>
> This honestly must be the first time I've seen someone on this list
> say "SATA" and "port multiplier" to conclude that such is the "more
> reliable solution" -- and in comparison to a SAS-based solution, no
> less.



I'm reasonably sure I've said words to this effect before on this list.


Usually there tends to be some grief about SATA+PMPs being prone
> to bus resets, timeouts and whatnot. Did I take the wrong medication
> this morning, or is Santa keeping a special eye on you Gordan?
>

Yes, sometimes you get bus resets when the disk starts to die, but you
get that in most cases anyway. It doesn't mean you lose all the disks on
that multiplier. And most SAS cards with more than 8 ports have built in
SAS expanders, e.g. a 16-port SAS card is often a 4-port card with
4x 4-port expanders built in (expander being the SAS term for PmP).

On the whole I've had orders of magnitude more problems with various
denominations of SAS cards (LSI, 3ware and Adaptec) than with simple
SATA cards with PmPs. I find that 4-port Marvell SATA cards with Silicon
Image PmPs is a particularly good combination for spinning rust.


>
> Of course, with a JBOD pool, if you lose one drive you lose the entire
> pool _anyway_, so...
>

I was referring to the SAS card in JBOD mode (as opposed to RAID mode).
I generally use suitably redundant storage for anything more important than
/var/tmp.

To unsubscribe from this group and stop receiving emails from it, send an email to zfs-discuss+unsubscribe at zfsonlinux.org.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://list.zfsonlinux.org/pipermail/zfs-discuss/attachments/20141223/f5989532/attachment.html>


More information about the zfs-discuss mailing list