[DRBD-user] concept behind drbd

Tim Jackson lists at timj.co.uk
Mon Jul 4 08:35:35 CEST 2005

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


On Sun, 3 Jul 2005 09:37:17 -0700
Curtis Vaughan <curtis at npc-usa.com> wrote:

> 2 computers w/postfix & courier-imap.
> Computer 1: has something like an 140 Gig HD.
> Computer 2: has something like eight 9 Gig HDs.

> Computer 1 will be the primary.

In the context of DRBD, "Primary" is a state, rather than a
specific machine. (That's not to say you can't have a preferred machine
if your cluster manager - e.g. heartbeat, or a human - supports it, but
generally speaking similar configurations are desirable and make sense
- after all, the nodes should both be able to perform the same tasks.)

> So it's my understanding that in order for drbd to work, I will have  
> to take the 140 HD on the primary and divide it up into 9 Gig  
> equivalents of the HDs on the secondary.

That would be one possibility, yes.

> So let's say say on the secondary I put say /home on /dev/hda2, /var  
> on /dev/hda3, root on /dev/hda1 - just for the sake of argument and  
> these are the drives that I want to be made redundant. Therefore,
> for each of these drives I will have to have an equivalent partition
> on the primary. And the partitions should be pretty much the exact
> same size.

The block devices (drives, partitions, LVM volumes...) that you are
replicating across the two nodes should match in size, yes. However, if
you have eight drives in a machine then you should probably be thinking
about some kind of RAID controller? Alternatively (or as well), I
would strongly recommend using LVM to allow you to deal with the
space more flexibly. (LVM lets you aggregate space from a number
of physical devices and slice it up in different ways, including
resizing/adding space if necessary). You can run DRBD on top of LVM
devices.

Tim



More information about the drbd-user mailing list