[DRBD-user] One big drbd or several smaller ones ?

Brian Candler B.Candler at pobox.com
Fri Jun 20 09:29:55 CEST 2008

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


On Thu, Jun 19, 2008 at 08:26:08PM +0100, Lee Christie wrote:
> Plan B
> ------
> Take the 11TB array (/dev/sdb) and create a single partition on it.
> Create a single VolumeGroup from /dev/sdb1.
> Create 8 LogicalVolumes from the VolumeGroup
> Use drbd to replicate each LogVol onto the second node.
> Format each /dev/drbd[x] with reiserfs
> 
> Advantages :
> + if drbd were to corrupt or suffer issue, perhaps that issue would be
> limited to one partition ?

+ you can move each resource's primary independently between the two
servers. This gives you the possibility of "live-live" operation, i.e. some
volumes served from one side, and some from the other, and to migrate them.

Whether this is useful to you or not depends on your application. It's
particularly useful if sharing volumes via iSCSI or NFS, or using them as
backing for virtual machines.

Aside: make sure you're happy with your decision to use reiserfs. My
employer suffered catastrophic data loss with reiserfs before, and doesn't
use it any more. You do also realise that the author may be somewhat
difficult to contact? :-)

But I don't want to restart the old "reiserfs vs everything else" debate. I
know there are plenty of people who use reiserfs happily who haven't
suffered data loss, presumably including you. Just be aware that there *are*
reports from a significant minority who have, and you should consider this
in your risk/benefit analysis. If your application's workload profile is one
that reiserfs performs very well on, this may outweigh the possibility of
corruption. Also, given that you have 8 volumes, restoring 1/8th of your
data may be acceptable.

>From the other side: although ext3 is robust, you should be aware that LVM
doesn't implement write barriers, negating the ability of ext3's journal to
prevent filesystem corruption. There is recent work underway to improve this
situation, but if you're using a stock distro kernel I don't think it will
be in there. See http://lwn.net/Articles/283161/ - there's a link from there
to a repeatable way to demonstrate corruption. Also
http://bugzilla.kernel.org/show_bug.cgi?id=9554

The lack of write barriers might also impact the integrity of reiserfs on
LVM - or it might be designed to cope without them - I really don't know.

Regards,

Brian.



More information about the drbd-user mailing list