Gluster on top of RAID5

[andrea68]

I was thinking to save disk and avoid 3 replication Gluster waste of space, to create a gluster distributed volume (with no redundancy) on top on Raid5 hardware hotswappable on 3 nodes.

It is a viable solution or a very bad scenario?

Tnx in advance...!!!

 

[Alwin]

Just for a thought, what happens if one node dies completely? I assume, then that portion of the data is not available and I guess the VM/CT using that data will fail.

 

[andrea68]

Just for a thought, what happens if one node dies completely? I assume, then that portion of the data is not available and I guess the VM/CT using that data will fail.

Yes, you are right.
Obviously the 3 nodes will be redundant in every other aspect to avoid or minimize fails.
But if a stop of services are allowed, data are safe in raid5 so you can recover from a node fail simply restarting it o reinstalling it.
This is good only on a small home project or where there's no SLA etc.

I think at this solution looking gluster docs:

https://docs.gluster.org/en/latest/Administrator Guide/Setting Up Volumes/

"You can use distributed volumes where the requirement is to scale storage and the redundancy is either not important or is provided by other hardware/software layers"...

If someone has another interesting solution with 3 nodes, local storage and Gluster (or Ceph) without losing 2/3 of disk space and have a decency avaibility, I'm all ears

 

[guletz]

Hi,

If someone has another interesting solution with 3 nodes, local storage and Gluster or Ceph without losing 2/3 of disk space and have a decency avaibility, I'm all ears

Maybe it will be better to use lizardFS. Using something like RAID5(lizardfs) you will lose only 1/3 of disk space!

 

[andrea68]

Never heard before...! I will looking into it.

 

[guletz]

Never heard before...! I will looking into it.

It will works better then glusterFS!