[Gllug] Swap Restrictions

[Grzegorz Jaskiewicz]

On 2006-04-12, at 10:35, Steve Nelson wrote:

> Hello All,
>
> I was always under the impression that on x86 architecture there is a
> 2G limit of size of swap areas.  I read this in the mkswap manual
> page:
>
> """ The  maximum  useful  size of a swap area depends on the
> architecture and kernel version.  It is roughly 2GiB on i386, PPC,
> m68k, ARM, 1GiB on sparc, 512MiB on mips, 128GiB on alpha and 3TiB on
> sparc64. For kernels after 2.3.3 there is no that limitation."""
>
> I had wondered if this still applied, given the "after 2.3.3"  
> statement.
>
>
Might depend on the kernel configuration it self also. I am just  
testing it now on sparc macine, will see. One thing, I am using 64bit  
kernel, with 32bit userland on sparc.
I hope you used mkswap -v1 /dev/some1 :) But as far as I remember,  
2.6 doesn't work with previous versions of swap structure. So mkswap  
must be using the newest version by default.

>
> So do I take it that the 2G limit is still in place?  If so how should
> we read the 2.3.3 comment in the man page?
>
> Finally, I have just built another two machines, both with 8G of RAM.
> The client requested 16G of swap(!).  There weren't enough partitions
> to make 8 x 2G partitions, and surely this is overkill anyway, even
> for a very very busy Oracle server?  I created 3 x 2G.  Just wondered
> what we think about such large swap partitions, and how they should be
> created?  More extended partitions and multiples of 2G?

You should use spare disc exclusively for swap than, remember that in  
2.6 you can create file swaps, they are (should?) be as fast as  
partition swaps. And so much easier to handle.

-- 

Grzegorz Jaskiewicz

"Better to remain silent and be thought a fool than to speak out and  
remove all doubt." - A. Lincoln





-- 
Gllug mailing list  -  Gllug at gllug.org.uk
http://lists.gllug.org.uk/mailman/listinfo/gllug