On Jul 14, 2010, at 6:17 PM, Jim Kusznir wrote: > Ok, my existing share is NOT LVM. Is it possible to convert a non-LVM > into an LVM without backing up the data? (I don't have anywhere big > enough to save the data to for a backup -- my new storage is only half > the size of my old storage).
Hi Jim, From a quick online search, it looks like converting is possible, but requires that you have more free space than used space on your partition, and even then, a lot of steps are required to shrink your partition and put LVM on the remaining space, then copy data over, and finally grow the LVM partition. You might consider this alternative: Ti Leggett wrote a script a while back that does a poor man's rebalancing by copying files from the filesystem back in so that the files are distributed evenly. Files get copied from smallest to largest, so you don't need to have new space larger than your largest file. I've attached his script. One nice thing about this approach is that if something goes wrong during the process, you haven't lost any data. The downside is that it will probably take a while. A thread of his experiences using it is here: http://www.mail-archive.com/[email protected]/msg00808.html -sam
pvfs2-rebalance
Description: Binary data
> > On Wed, Jul 14, 2010 at 2:11 PM, Emmanuel Florac <[email protected]> > wrote: >> Le Wed, 14 Jul 2010 09:42:26 -0700 vous écriviez: >> >>> Question is: >>> 1) should I try and expand the existing RAID6 volume (over 15 740GB >>> disks presently), or just add a second volume? >> >> It's risky to expand a single volume across several physical device. >> >>> 2) If I add a second volume, how best do I tell pvfs2 to use this? Or >>> should I try and LVM it with the old volume? (Can I even do that?) >> >> Yes, definitely use LVM to aggregate the various volume. Don't stripe >> across volumes, so that an unavailable/failed device doesn't fail the >> whole setup. >> >>> 3) Do I need to mess with the MetaHnadle ranges? (The amount of space >>> will be the same on all 3 servers) >> >> If you simply extend existing PVFS2 underlying volumes, you don't need >> to change anything. >> >>> 4) is there any manual rebalancing I need to do? >> >> If you create new pvfs storage space (not recommended), you'd need to >> scan all files, run pvfs2-stat on them, and defragment them if their >> number of datafiles is less than the number of pvfs2 devices. >> >> That's precisely why it's better to simply expand the existing >> volumes... :) >> >> -- >> ------------------------------------------------------------------------ >> Emmanuel Florac | Direction technique >> | Intellique >> | <[email protected]> >> | +33 1 78 94 84 02 >> ------------------------------------------------------------------------ >> > > _______________________________________________ > Pvfs2-users mailing list > [email protected] > http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users
_______________________________________________ Pvfs2-users mailing list [email protected] http://www.beowulf-underground.org/mailman/listinfo/pvfs2-users
