Dan Lund wrote:
> Well, here's the whole setup:
>
> An AMD64 Opteron 246 w/2 gb ram, with 2 IDE "system" disks installed.
> I raid1 them through software raid.
>
> I have 16 SATA disks attached via an Areca SATA RAID card which is
> configured with a RAID6+2 failover setup.
> I use LVM2 to split them up amongst their different partitions, and I
> have 3.2TB allocated to this particular partition.
>
> After looking at the raw /dev/sda disk (allocated by the Areca driver,
> and setup as a physical device in my volume group) it is labelled
> msdos. Would that cause a problem being so low in the layers?
Yes, an msdos label will cause this problem.
msdos labels report inaccurate info to the OS about your device/partitions.
Thus your "disk full" issue when you should have had additional physical
space available. As soon as you get a gpt label on their you'll get
access to all your space minus journaling space requirements.
>
> If I change it, how would I make it accessible again via parted? I'm
> not terribly versed in parted. Just looking at the params, my guess
> would be to run mklabel, and once the partition table is killed, run
> rescue with a start and end.. but where to get those numbers?
>
You'll want to run mklabel and set up a gpt label.
At that point you'll see that there are no longer any partitions
configured for the device. From there I've always created new
partitions (`mkpart`) and reloaded given that in my situation, a failed
recovery would have cost me far more time and heartache than actually
dumping and reloading several terabytes of data.
That said, my hypothesis is that you can get the numbers you need from
the output of the parted print command (start/end).
That will likely be useful in bringing your partitions back to life
(`rescue`) -as they exist right now- which of course is the wrong size.
At that point you would want to `resize` the partition up to the max (0%
to 100% works) size assuming you don't have another partition further
out on the platter than the one you're working with.
> Either way, I'm dumping the whole partition to tape tonight.
>
Probably a good call.
>
>
> On 5/2/07, steve crandell <steven.crandell@icrossing.com> wrote:
>> Take a gander at your disk label.
>> # parted /dev/sdx
>> > print
>>
>> I'm guessing it will be msdos.
>> If so, that's your problem.
>> Consider using gpt instead.
>>
>> Note: changing a disk label will render your data unusable, though it
>> will technically be `rescue`-able using gnu parted.
>>
>> That said I'd consider copying it off-disk prior to making any changes
>> if possible.
>>
>> Dan Lund wrote:
>>> Well I doubt its failing hardware on a raid6+2, I was leaning more
>>> towards device constraints and was wondering if anyone else has dealt
>>> with a device greater than 2tb.
>>>
>
--
Steve Crandell
Linux System Operations
480.282.6047 direct
480.239.9286 mobile
480.505.5801 fax
steven.crandell@icrossing.com
www.icrossing.com
ATLANTA | CHICAGO | DALLAS | NEW YORK | SAN FRANCISCO | SCOTTSDALE
Winner of OMMA Magazine's 2005 Agency of the Year: Best Search
---------------------------------------------------
PLUG-discuss mailing list -
PLUG-discuss@lists.plug.phoenix.az.us
To subscribe, unsubscribe, or to change you mail settings:
http://lists.PLUG.phoenix.az.us/mailman/listinfo/plug-discuss
(text/plain)