[NCLUG] 1TB Blocksize Limit?

Andrew Patterson andrew at fc.hp.com
Wed Aug 28 17:38:42 MDT 2002

On Wed, 2002-08-28 at 17:22, DJ Eshelman wrote:
> Okay, Let's just jump into it since I can't think of anything clever.
> My RedHat Advanced 
> Server troubles continue!  Yay!
> We finally found a FibreChannel card that works out of 
> the box (pretty much), the QLogic QLA2310-CK 2Gbit over Copper HBA.  This was after not-so-
> valiant efforts by Adaptec and the best wishes and demos from LSI (sorry guys, but it just 
> don't work yet!)
> So yay, I get the drivers loaded, everything looks good until I try to find 
> the actual storage device (it's an Arena Indy 2600-FC chassis- a device that houses 16 120GB IDE 
> drives and puts them into a RAID 5 Array (or whatever, there's a lot of optoins) with two Fibre 
> Channel (2Gbit Copper) interfaces.
> I dig a little deeper and it appears that the total size 
> of the array is too large (1.6TB) for Linux.  This is news to me, so I call them up and confer with a 
> few engineers.  We know the device works because when I split the drives into two 800GB 
> 'virtual' drives it sees them both, it apparently is a limitation of the Kernel in regards to 
> blocksizes.  Apparently anything over 1TB is not supported by the Kernel I'm using (Advanced 
> Server is up to 2.4.9-e.8).  Now, I'm an ignoramous when it comes to Kernel stuff, I will admit- 
> when the tech started talking about 31 pointer bits per block number or whatever, I started 
> drooling and reaching for my club to go hunt for dinner.  But this doesn't help matters since 
> our goal was to have all 1.6TB of data usable all in one drive mount.
> So does anyone out there 
> have any ideas for me?  I'm about at wits end with this project and I swear my mind isn't working 
> any more- it just flat gave up.
> I can find plenty of 'data' on the web as far as XFS and 
> filesystem limits, but according to RedHat this is a block device problem, not a filesystem 
> problem.  The other 'gotcha' is that I can't really do any major modification to the Kernel and 
> still get support.  At this point I'm about ready to say screw it, and go with whatever will get 
> the job done.  Any ideas out there?
> Keep in mind that this is an Intel Xeon (x86) system, so the 
> obvious answers are moot (get a 64 bit platform, etc).

I can use 2 TB LUN's on a kernel running 2.4.17 running XFS. I believe
it should work on all 2.4.x kernels.  What is failing?  Partitioning the
drive, putting a file-system on it?  What does /proc/partitions say?


> I appreciate it.
> -DJ 
> Eshelman
> dj at sgc-inc.net
> -------------------
> _______________________________________________
> NCLUG mailing list       NCLUG at nclug.org
> To unsubscribe, subscribe, or modify your settings, go to:
> http://www.nclug.org/mailman/listinfo/nclug
   Andrew Patterson                          Voice:  (970) 898-3261
   Hewlett-Packard Company                   Email:  andrew at fc.hp.com

More information about the NCLUG mailing list