Re: RAC storage newbee question
- From: Justin Mungal <justin@xxxxxxx>
- To: ahmusch@xxxxxxxxx
- Date: Mon, 26 May 2014 20:26:02 -0500
"I'm told by the storage admin that he can not provide raw disk. Although
he could provide raid 0."
Adam, it sounds more like the storage admin is saying he can't present a
LUN that is not in some type of RAID.
Chris, at work all of our LUNs are some type of RAID. Generally they are
RAID 10 because of its performance and redundancy benefits. We generally
create LUNs that are sized in increments that match the increment size that
the customer will want to expand by in the future (or a reasonable
prediction of such). So let's say we go with a 400GB +DATA diskgroup, a
200GB +FRA diskgroup, and the customer wants to expand in 100GB increments
on both diskgroups. The LUNs would be configured as follows:
4x 100GB RAID10 LUNS (which would use double that amount of raw storage),
with external redundancy
2x 100GB RAID10 LUNS, with external redundancy
The diskgroup which will store the voting disks and OCR is another story.
Since it requires so little storage, and is critical to cluster operation,
we configure it as follows:
5x 1GB RAID10 LUNs, with high redundancy
This is an "extra safe" option that doesn't cost us much in terms of
storage.
In the future, we can present new 100GB LUNs to the server, and add them to
the diskgroup online. If the customer for whatever reason wants to change
the storage layout completely, we present the new LUNs to the server, and
do the add/drop operation with a single command in order to reduce the
amount of extent rebalancing.
As Mark mentioned earlier, it's hard to beat hardware RAID. Normal/High
redundancy makes more sense to me in JBOD situations.
Hope this helps!
-Justin
On Mon, May 26, 2014 at 7:56 PM, Adam Musch <ahmusch@xxxxxxxxx> wrote:
> I would be very surprised to hear of a storage admin that can't provide
> RAID 1 volumes as raw disk - then you can add each RAID 1 volume as an
> externally redundant ASM disk, and then stripe across them.
>
>
> On Mon, May 26, 2014 at 11:40 AM, Chris King <ckaj111@xxxxxxxx> wrote:
>
>> Thanks in advance for your feedback on this question.
>>
>> We built RAC in a lab using VMs and now are planning to build on metal.
>> We're using ASM for disk storage management, so I requested raw disk.
>>
>> I'm told by the storage admin that he can not provide raw disk. Although
>> he could provide raid 0.
>>
>> So my question is:
>> If I ask the storage manager to provide disk that is already raid 0+1,
>> and then I create ASM disk groups using the "external" option. Is this
>> approach any better/worse than providing ASM raw disk to manage?
>>
>> (I would use the external option so that I don't end up striping and
>> mirroring disk that is already striped and mirrored.)
>>
>> Or would it be better to have him provide all the disks as raid0, and
>> have ASM handle all the redundancy?
>>
>>
>>
>>
>>
>>
>>
>
>
> --
> Adam Musch
> ahmusch@xxxxxxxxx
>
Other related posts:
