[OpenIndiana-discuss] Harddisk > 2TB

[Sašo Kiselkov]

On 01/20/2013 04:56 PM, Sašo Kiselkov wrote:
> On 01/20/2013 04:40 PM, Ulrich Hagen wrote:
>> Jim Klimov wrote:
>>
>>> On 2013-01-20 13:57, Ulrich Hagen wrote:
>>>> Hello everyone,
>>>>
>>>> I have recently added an Intel SASUC8I controller to my home file
>>>> server, and hooked up eight Western Digital Red 3000GB disks to it.
>>>>
>>>> Only after I started filling up the new pool I noticed that the
>>>> capacity of the pool is not what I expected it to be.
>>>> And really: 'format' lists e. g. one of these disks as
>>>>    c1t0d0 <ATA-WDC WD30EFRX-68A-0A80-2.00TB>
>>>
>>> As a couple more wild guesses:
>>>
>>> 1) How does (the HBA's) BIOS recognize them? Does it confirm that
>>> the disks are 3Tb and not mis-marked on paper label? ;)
>>
>> Now that I looked more closely at the boot (BIOS) messages, I have
>> noticed that the controller lists these disks as having a capacity of
>> 2047 GB. Also a (too late) search in the internet revealed that 2TB is
>> the limit of this controller.
> 
> It indeed appears to be a hard limit of the LSI SAS 1068e chip, no newer
> firmware appears to fix this issue (which is bizarre, but I suppose LSI
> also knows how to force customers to upgrade). I suggest you pick up any
> one of the widely available LSI SAS 2008-based cards out there - in
> general, OEM cards tend to be cheaper than LSI-original ones, despite
> running the same hardware and being reflashable to LSI firmware. See
> http://www.servethehome.com/lsi-sas-2008-raid-controller-hba-information/ for
> a list of suitable candidates. I personally prefer Dell's PERC H200 -
> essentially a pure LSI 9211-8i, no reflashing needed, JBOD support and
> runs like a champ under OI.
> 
>>> 2) Is it possible that for some reason these disks use an MBR
>>> partitioning table instead of GPT? The former would max out at 2Tb.
>>
>> They were and are not partitioned. I took them out of their boxes,
>> connected them to the controller, started OI and created the pool
>> using entire disks.
>>
>> And, to reply to Sašo Kiselkov:
>> ashift is 9, these disk lie about their native sector size. So my pool
>> will never be as fast as it could.
> 
> Nope, they don't. What you're hitting is a bug in ZFS which incorrectly
> handles Advanced Format drives. I have the same kind of drive with the
> same formatting and my pools are ashift=13, because I created them with
> the patched zpool command from Illumos source. If possible, I recommend
> you re-create your pool with the correct ashift - it is possible.

FYI: here's the relevant link:
http://wiki.openindiana.org/pages/viewpage.action?pageId=4883847

(and minor correction of myself: 4k sectors is ashift=12, not ashift=13)

Cheers,
--
Saso