1

Imagine I have two servers, each server has two disks in RAID 1. Now I merge the two arrays with iSCSI to one RAID 1 disk.

Two questions: Can I do the merging of the 4 disks in one go? I can't image how. First I will have to install the OS, and then the RAID controller is already set up to RAID 1.

If a whole server fails the other server would continue working without any problems? Does iSCSI notice that the other server is missing and treat this as if the two disks were broken? When the server comes back online the data is resynced, as if I installed new disks into a array?

Is this even possible?

Improve this question

3 Answers 3

Reset to default
2

I think what you are looking for is DRDB (Distributed Replicated Block Device) http://www.drbd.org/. It allows you to replicate one volume (whether its a normal partition, an LVM volume, RAID or LVM on RAID) to another server.

It is often described as a network equivalent of RAID1.

DRBD® works on top of block devices, i.e., hard disk partitions or LVM's logical volumes. It mirrors each data block that it is written to disk to the peer node.

DRDB diagram, www.drbd.org/home/what-is-drbd

Improve this answer
1

I don't think a raid is what you should use between the systems. With a software raid it should work to set up the configuration. But you MUST NOT use the storage on both systems at the same time. This will corrupt your data!

If you are under linux, have a look at drbd. This is designed to mirror storage across a network with master/slave scenario.

Improve this answer
0

In this scenario I imagine you would use software RAID to create a RAID 1 array across 4 disks, however this seems like an extreme waste as you would effectively only get 1 disk of 4:

If you have 4 250GB disks split across two servers each with their own RAID 1 set you then end up with two 250GB arrays. If you then mirror them a second time you're down to a single 250GB disk from 1TB, which is pretty bad.

Unfortunately Software RAID only supports RAID 1 for boot, so this would be the only working scenario.

If you wanted to, however, this should work.

If your goal is to maintain redundancy should one of your iSCSI targets or one of your disks fail I would recommend using RAID 0 on both targets, this way you have redundancy for 1 disk or one target failure which, for a 4 disk deployment, should be enough. This way you would have a mirrored 500GB array (if each of your disks is 250 of course).

Improve this answer
3
  • Not true. You'd effectively have something like a RAID10, with a usable capacity of 2*n out of of 4*n (where n is the size of your disks)."Software RAID only supports RAID 1 for boot" is false, too. For example you can boot whatever RAID if you have a non-RAID Partition.
    – Craig
    Jun 28, 2011 at 10:48
  • Firstly, mirroring an already mirrored disk would create 1/4 of the storage: (((250GB*2)/2)*2)/2. Secondly if you have a non-raid partition then it's not an issue is it? If you want to provide redundancy for all partitions using a raid, however, RAID 1 is the only option.
    – Alex Berry
    Jun 28, 2011 at 11:35
  • No, still wrong. Say you have 6 disks. Think of it like this: there are two sides, 'original' and 'mirror'. Each side has 3 disks that hold the same content as the other side (thus 'mirror'). So the total amount of storage is 3*disksize here, which is n*disksize/2 where n is the number of disks you have. Read en.wikipedia.org/wiki/RAID_10#RAID_10_.28RAID_1.2B0.29
    – Craig
    Jul 7, 2011 at 18:07

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .