Re: SAS and SATA (arrays) on one controller (LSISAS1068)?



"Eric Moore" <ericmooreco@xxxxxxxxxxxx> wrote in message news:AtmdnZjAVtbhncjbnZ2dnUVZ_jadnZ2d@xxxxxxxxxxxx
"Folkert Rienstra" <see_reply-to@xxxxxxxx> wrote in message news:46536f64$0$97234>>
I'll use FreeBSD's mpt fusion driver, thus I guess
two raid 1 arrays, one 2x SCSI and the other 2x
SATA, should work together well...

The warning I mentioned is indeed found in the
Embedded MegaRAID Software User's Guide Ver 1.0
(page 1-4).

So this is a concealed warning that their software is broken,
as originally suggested.

Folkert - This is what I found out.

Gee wiz, no kidding.

The reason why RAID volumes can't be mixed is due to the way bad block management is
implemented between SCSI(SAS) and SATA is different.

Nope, it's not. They both have it.
The difference is that you can configure it on SCSI (including disabling
it) but not on (S)ATA.

For example, in SCSI standards, there is automatic reassignment of bad blocks when an
bad sector is encountered,

Wrong. You got it backwards. That's (S)ATA.
SCSI bad block management can be configured where (S)ATA can not (automatic).
So SCSI's auto bad block management can be switched off where SATA's can not.
In SCSI, near bad blocks can be forcibly reassigned where in (S)ATA it can not.

whereas with SATA this is not specified in the standards,

Because there is nothing to specify, currently.
Maybe in future standards the need may grow for configurability of it.
Or add forced reassign.

so firmware has to manually handle this by allocating space at the end of the drive for bad block table and free space for
available sectors.

That is utter crap.
That is what the antiquated RAID systems do for standard and which fails with SATA.
(Failure being relative when a perfectly good 'bad' block is set aside when it need not).

I guess

Exactly!

there would be issues when there is a mirror, and a bad sector was encounterred when one drive was SATA, and the mirror was SAS..

Not a problem if both do automatic reassignment. But not a problem either
if they don't. Just different ways of handinge it.
Auto bad block management didn't make a problem either with previous
methods of setting away bad sectors/clusters in filemanagement systems.
It can work perfectly side by side.
It has to because automatic reassignment doesn't work in all situations sin-
ce it works differently for reads and writes. It's not a perfect cure for all.

Bad Block Management would be spelled out in the SCSI(SAS) standards are on www.t10.org.

Let me know if you need me to point out which spec to read.

No thanks. But feel free to read them yourself sometimes.

The SATA standards can be found on www.t13.org, and https://www.sata-io.org/.

In addition, another reason, is we were forced to not mix SAS/SATA in RAID
volumes from some of our big box customers, such as HP, Dell, and IBM.

And I can easily see why.


I found a confirming information in the
"TECHNICAL MANUAL LSISAS1068 PCI-X to 8-Port
Serial Attached SCSI/SATA Controller" (Oct. 2005,
Ver. 2.1) that the LSISAS1068 "allows mixed
connections to SAS or SATA targets" (page 1-10).

Can these people at LSI not even write one sentence without starting
another confusion? : 'mixed' and 'or' are contradictory.


As long as the end devices are not members of the RAID volume, you can have a mix of SAS and SATA devices connected to the
controller. If your creating a RAID volume, they have to all the same.

The sentence is what it is: 'mixed' and 'or' are contradictory.

Is that confusing or contradictory?

Both, if the controller lets you mix them when it's not allowed.
Both, if the controller lets you mix them and an initially working combina-tion exists.
If not then the warning is obviously false since the situation cannot exist
and no harm is done.


Eric

.



Relevant Pages

  • Re: SATA vs. SCSI, RAID?
    ... There's a wealth of RAID information on the Adaptec site. ... page comparing SCSI and SATA RAID. ... with SATA in servers, so I'm not commenting. ... > However going from SATA drives with software RAID to SCSI ...
    (microsoft.public.windows.server.sbs)
  • Re: [OT] Thank God for the Guy who invented RAID
    ... Yes I can't wait Until SATA speed get's faster ... > gives only a modest performance advantage over SBS2003's intrinsic SW RAID ... > heavy continuous load. ... > own vs. two drive HW SCSI RAID 1. ...
    (microsoft.public.windows.server.sbs)
  • Re: Addendum
    ... BTW, why do you recommend a RAID 10? ... > I have found that real inexpensive SATA drives do not perform at the level ... > SCSI, but for a small work group and the uses you specified a good six ...
    (microsoft.public.windows.server.general)
  • Re: What still uses the block layer?
    ... often _only_ be used by going through the SCSI midlayer. ... That's because modern USB, ATAPI, SATA ... As far as I can tell the hard drives do not have serial numbers easily ...
    (Linux-Kernel)
  • Re: 3ware escalade 8006 SATA RAID recommendations; dual boot windows
    ... Doesn't command queing bring SATA much closer to SCSI performance? ... to another manufacturer's SATA RAID controller fail? ... both the disk drive and the adapter card are Frankenstein hardware. ...
    (Debian-User)