[DRBD-user] DRBD + ZFS

Robert Krig robert at bitcaster.de
Sat Nov 5 13:37:24 CET 2011

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

If you are setting up a storage Server, go with what you know and what
his been tried and tested.
A high volume storage system is not the place to be testing out unstable
or experimental implentations of filesystems.

Well, unless you don't mind backing everything up and reformatting
further down the line once you run into trouble.

Also, seriously consider what you actually need vs. what you want.

The simplest setup with DRBD is a primary>secondary node setup. You can
use a common filesystem like ext3/4 or XFS, and you have one node which
is active, and another which is a standby. Switching over can be done
manually in just a few quick commands.
If you need automatic failover you might want to check out heartbeat or
pacemaker in conjuction with drbd.

If you want both nodes to be available simultaeneously, then you have to
go primary/primary setup. Which also requires you to use a cluster aware
filesystem such as GFS or OCFS2. This adds another layer of complexity
to your setup. Especially if something goes wrong.

If you don't need fast random access throughput on tons of small files,
then you might want to take a look at GlusterFS. The advantage of
GlusterFS is that it works with pretty much any filesystem. You
designate a directory to be a glusterfs export. Then you mount that
export on a local or remote machine, and all file operations done
through that mountpoint get replicated/distributed to all your nodes.
If GlusterFS fails to work, you still have your files in the original
directory to work with.

However, GlusterFS is definately NOT recommended for serving a lot of
small files.

Anyway, food for thought.

More information about the drbd-user mailing list