4

We are storing files on a network share, our application is mountig this drive using nfs. Now we have the requirement to have this drive redundant.

  1. We can add a second server and then synchronize the direcory with this server. What are the best approaches to do the synchronisation?

  2. If one server is out of order, how can we do the failover? Is there a way in nfs to specify a backup drive and nfs is automatically getting the data from the backup drive? If this is not possible with nfs, is there another filesystem which supports something like that?

markus
  • 1,080
  • 5
  • 18
  • 38

3 Answers3

3

It looks like this is a good candidate for HA Cluster with DRB. Something like this:

http://www.howtoforge.com/highly-available-nfs-server-using-drbd-and-heartbeat-on-debian-5.0-lenny

Not the best description, though, so I would suggest going here as a starting point.

http://www.drbd.org/

Essentially, you have two servers and the drive is being replicated from server to another. that drive is exported via NFS, which is then mounted by the client via dedicated IP on one of the servers. If one server fails, , failover will be almost transparent, as it is just the matter of making the other server the primary and moving that IP from one server to another (manually or automatically).

Rilindo
  • 5,078
  • 5
  • 28
  • 46
  • Do you have personal experience with this software? – markus Sep 15 '11 at 16:59
  • I worked with DRBD at my previous position. It worked well for replication, but due to time and legacy configuration, I was not able to implement the HA portion. Putting aside that, it shouldn't be that difficult if you do it right. :) – Rilindo Sep 15 '11 at 17:22
3

You can also try Glusterfs

http://www.gluster.org/

With glusterfs you can make different ways of replication. The servers can do it themselves, or the clients replicates it. We've been working with it for a few months now and it hasn't let us down (yet)

blauwblaatje
  • 953
  • 1
  • 6
  • 19
0

The easiest and long term most stable solution would be to use Netapp filers. Very fast failover, fully redundant, FlexCache cards for performance spikes etc, the arguments for Netapps are many but the initial capex price tag may look high until you realize what you actually get for your money.

HampusLi
  • 3,478
  • 17
  • 14
  • If he doesn't need all the features or capacity of a NetApp. If this is just for a few GB of shared storage for a webfarm for example, that probably would not be justifiable. – mfinni Sep 15 '11 at 15:59
  • So far the drive will never get larger than 50 GB. – markus Sep 15 '11 at 16:31