[DRBD-user] a Newbie's Questions

Corey Edwards tensai at zmonkey.org
Mon Oct 24 04:53:52 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 Mon, 2005-10-24 at 05:08 +0300, Louai Al-Awami wrote:
> I am new to DRBD and have some  questions.

Welcome!

>      1. I am using DRBD under UML, and I am not sure which block
>         device to use for the "disk" field for my resource. I use the
>         disk I found when running df command, but I got a message
>         saying "low level device already mounted". Here is my
>         drbd.conf:

It means just what it says. /dev/ubd/disc0/disc is mounted and DRBD
cannot access it until you unmount it. Once you turn on DRBD on that
device *never* mount it directly again. Bad Things(tm) will happen.
Use /dev/drbd/0 instead. Please make sure you understand the
difference. /dev/drbd/0 is mirrored, /dev/ubd/disc0/disc is not.

>      1. Is DRBD used only to synchronize the whole disk? Can not I
>         used to synchronize directories instead?

DRBD mirrors a block device. It knows nothing about files and
directories, just the raw block data that is used to represent them. So
the answer is no, you can't synchronize a directory. But, you can create
multiple partitions and synchronize those. So instead of one partition
at /home you could have /home/user1, /home/user2, /home/user3, etc. That
of course would have its own set of ramifications which you would
certain want to consider. LVM may also help you.

>      1. What is the meta-disk for? And can I use a meta-disk that is
>         smaller than 128MB?

Meta data basically keeps track of which parts of the data disk are in
sync. Here's an example. Say your secondary goes offline. The primary
continues running and keeps track of which blocks are changed so that
only those blocks have to be sent over to the secondary when it returns.
Let's say the world collapses and the primary goes down too. If that
list of changed blocks was only stored in RAM (as was the case in 0.6,
IIRC), the primary has to do a full sync (send all data) over to the
secondary. With meta data it can read the list from disk and send only
the differences it would have sent had it never gone offline. Make
sense? Meta data is a Good Thing(tm).

While we're on the subject, I'll recommend using external meta data.
Internal works just fine if you plan ahead and keep your head about you
at all times, but I can't say I always think straight at 2 am. Or in the
middle of the day, for that matter. I avoid internal meta data wherever
possible.

Corey

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20051023/dec82d45/attachment.pgp>


More information about the drbd-user mailing list