[DRBD-user] Implementing drbd with an existing filesystem

Lars Marowsky-Bree lmb at suse.de
Thu Mar 10 17:32:18 CET 2005


On 2005-03-10T10:15:02, kmoogle <kmoogle at gmail.com> wrote:

> I'm interested in using drbd, but cannot modify the active fileserver
> with customer data outside of perhaps replacing the kernel. No major
> (or potentially destructive) modifications can be made to the primary,
> active server.
> 
> Can I implement drdb without adversely affecting the primary
> fileserver? What changes would I need to make to the existing
> filesystems (if any) to either the primary or secondary in order to
> implement drbd?

You'd need to shrink the filesystem by 128 MB on the primary, or make
the underlaying physical storage 128MB larger. (or a bit more, if you
want to play it safe.)

The trick is to create a gap at the end of the device for drbd to store
it's metadata in.

Another alternative would be to create an additional 128MB partition on
the RAID and use drbd with it external metadata.

Then you could put drbd on top, and tell the secondary to sync from the
primary. (In an initial setup, they'll both come up secondary/secondary
with drbd; use "drbd invalidate_all" on the slave, and then use "drbd
primary" on the, well, the primary. Replication will begin.)

Depending on how comfortable you feel with this and how important your
data is, you might want to buy support from LINBIT and/or SUSE/Novell
for that one. (Of course, SUSE can only help you if you run SLES;
otherwise, Linbit will be glad to have your business ;-)

Having a backup ready is, however, never wrong advice.


Sincerely,
    Lars Marowsky-Brée <lmb at suse.de>

-- 
High Availability & Clustering
SUSE Labs, Research and Development
SUSE LINUX Products GmbH - A Novell Business




More information about the drbd-user mailing list