On Thursday 16 February 2006 17:21, Alexander Skwar wrote:
>
> > You *can* tell LVM where to put LVs but you do not *have* to.
>
> But how do you actually do that? Or are you talking about
> the "allocation policy"? Like "--contiguous y"?
>

Well, first of all, you can pass lvcreate a list of physical volumes that are 
then used to allocate extends for the newly created logical volume. By the 
order of LV creation, you determine the sequence of LVs on the PVs (or ore 
correctly, the sequence, in which the extends of one or more PVs are 
allocated to one or more LVs).

Then, you may use lvmove to move a LV to another PV. You may use lvsplit to 
split a LV into two or more parts and then use lvmove to move these part-LVs 
around.

Thirdly, you can (either by hand or by using a more sophisticated tool like 
EVMS) alter the mapping of LV extends to PV extends.

There are surely even more ways to tell LVM where to store LVs, but these are 
the ones that come immediately to my mind.

Kind regads
Martin
-- 
Dipl. Wirtsch.Inf. (Univ.) Martin Eisenhardt

Otto-Friedrich-University Bamberg
Department Business Informatics and Applied Computer Science
Media Informatics Group

D - 96045 Bamberg

fon: +49 (951) 863 2856
fax: +49 (951) 863 2852

www: http://www.mneisen.org