4

My company has only a single storage box (SuperMicro hardware + chassis with a LSI raid card and 12 SAS 7.2k drive) in its current setup. It runs on linux and we hook it up to our fileserver via iSCSI. We can't afford a high-end SAN. While it has redundant parts in it (dual PSUs, RAID 10), I would like to add in a secondary box for replication purpose.

I'm considering the following two options to replicate the data from node 1 to node 2

  1. Using DRBD to create a network based mirror. Active/passive. Node 1 remains as iSCSI target, and setup drbd underneath it so that data can be replicated to node 2. If node 1 goes down due to networking problem, I suppose the recovery is much faster since we don't have to resync the data from node 2 from scratch once node 1 is plugged back into the network.

  2. Setup both nodes as an independent box. Each of them will expose a LUN to our MS server 2008 FS. From there, I will setup a windows based software mirroring using both LUNs. My concern is that, in the event one of the LUNs drops out, we will have to rebuild the entire array, which will be very time consuming as we have about 6TB of data to copy over.

To me, both serve the purpose well, but the DRBD option seems more complex to manage. Please advise on the pros and cons.

Scott Pack
  • 14,907
  • 10
  • 53
  • 83
gostan
  • 53
  • 1
  • 3
  • 2
    Is a Windows file server your only client for this storage server? If so, back up your data and install Windows on it directly. For redundancy, install windows on the second server and use Windows DFS for automatic load balancing and fail-over. Extra hardware and software to add a "SAN" layer would seem to be over-kill. If Windows isn't your only client, then drbd is your best option for replication since it's more complicated to have each Windows client mirror its LUNs. – JakePaulus Oct 29 '11 at 22:39

2 Answers2

6

I would use a combination of DRDB and Heartbeat to make a HA iSCSI solution out of your situation. You can add Heartbeat to the DRDB setup to have the failover happen transparently to the file server.

MDMarra
  • 100,734
  • 32
  • 197
  • 329
0

We're using GlusterFS for our SAN. It is a software-only solution that runs in userspace. If you can install Gluster's client software on the clients (e.g. they're GNU/Linux), you might want to give it a try. If you're dependent on iSCSI, it might not be the solution for you, though.

Daniel
  • 3,047
  • 5
  • 22
  • 27