[sclug] LVM & 2Tb limit

[Martin Summers]

hello !


-----Original Message-----
From: sclug-bounces at sclug.org.uk [mailto:sclug-bounces at sclug.org.uk]On
Behalf Of Alan Pearson
Sent: Tuesday, February 15, 2005 5:19 PM
Cc: sclug at sclug.org.uk
Subject: Re: [sclug] LVM & 2Tb limit


On 15 Feb 2005, at 16:48, Martin Summers wrote:

> Hello,
>
> Redhat Enterprise tends to have 2.6 kernel functionality back ported 
> into 2.4, so you *should* be OK. It would be worth checking in the 
> docs for confirmation on this as I am only 50 % sure. The largest I 
> have ever done is 1.5 TB on Linux - this worked fine until it went 
> wrong one day...
>


I've found some info which says that RHEL 3 does NOT support block 
devices / filesystems larger than 2TB

http://www.arnnet.com.au/index.php/id;403016224;fp;512;fpid;47886160


(MS) - doh ! OK, fair enough. 

> My experience of large single volumes on linux are that if you ever 
> have a bad power failure or a situation where the machine crashes 
> badly, having to fix the filesystem will take forever (regardless of 
> if it is journaled or not) It took me 3 days non stop to fix my 
> filesystem on 1.5 TB, and even then, I had corrupt files. Restoring 
> from backup was the only thing to do. On a filesystem that size, it 
> may not be feasible to fsck it in a "reasonable" amount of time 
> (depending on what you call "reasonable". If you can break it down 
> into smaller partitions, you will be glad of it if you are ever in 
> this unfortunate positon (I speak from experience here !)

Interesting, what FS was it ?

(MS)It was ext3 - but the RAID controller only had 128MB cache, and it died mid-write. I did a series of tests on transferring large and small files, while simulating a power crash on a test RHEL 3.0 with logical volumes. My conclusion is that you will always get a certain percentage of corrupt files, depending on what was writing at the time the power goes.
It is easy to fix the filesystem, but knowing if the files are corrupt (truncated) or not afterwards is not so easy....



> Also, backing up a filesystem that large is going to take time......

Backups, we don't need no stink'in backups !!!
Seriously !!!
The content is dynamic and changes regularly.
Also, it's sitting on multiple HW Raid devices.

(MS) - yes, so was ours! Power can do some nasty things (ours arc'ed and fried everything in that area....I can still remember the smell of smoke......)

So I can take around 8 disk failures (in the right places).

(MS) <nod> we took 5 at the time of our power arc...

>
> I guess that's my two-penneth worth ;-)
>
> -Martin Summers

-Martin
_______________________________________________
sclug mailing list
sclug at sclug.org.uk
http://www.sclug.org.uk/mailman/listinfo/sclug