Hi,

Thanks. I'll try to change that yes, but as I said, it's been working fine for all other SL distributions, including SL6...

Regards

-----Message d'origine-----
De : Steven J. Yellin [mailto:[log in to unmask]] 
Envoyé : mercredi 21 novembre 2012 19:01
À : SCHAER Frederic
Cc : [log in to unmask]
Objet : Re: SL5.8, anaconda, pre-exec scripts, and swap issue

     When swap partitions are created with fdisk, there's a "t" option that 
is used to set the partition type.  If swap partitions are created with 
mkpart in parted, the "FS-TYPE" can be included in the command.  I'll bet 
if you change the scriplet to create the swap partitions with the Linux 
swap partition type, the boot software will then see the type and know to 
do the right thing.

Steven Yellin

On Wed, 21 Nov 2012, SCHAER Frederic wrote:

> Hi,
>
> We're deploying our nodes using pre-exec scriptlets, in which we do our own partitioning.
> In this scriptlets, we have things like that :
>
> 1-      Create swap paritions
> 2-      "mkswap those partitions"
>
> The issue is the following : until now, everything was working fine.
> With SL5.8, at reboot, our nodes only have 2GB of swap enabled per swap partition.
>
> If we do:
>
> -          swapoff partition
> -          mkswap partition
> -          swapon partition
>
> Then the swap size appears to be what it should be (up to 128GB depending on the node, so 2GB clearly is a not a good thing for
us).
>
> Off course, we could work around this using several methods, but that would be work arounds only.
> I tried installing a node using anaconda's partitioning and no pre-exec scripts : then the swap is correctly defined/used.
>
> Finally, I noted that on the failing systems (pre-exec partitioning), the partition type is 83 (linux), while on the anaconda-made
> swap, partition type is 8e (Linux swap / Solaris)
>
> Bug or not a bug ?
>
> Thanks
>