If theres a need to share a volume between multiples nodes of a Scaleio cluster, for example by a active-active samba instance, whats is supposed to be the filesystem that the ScaleIO volumes must me initialized? GFS? OCFS? Redhat GFS? It must be a shared filesystem BUT NOT a clustered one right? GlusterFS is not a option I think because this FS is competing in funcionality with ScaleIO? or GlusterFS CAN be used on top of ScaleIO?
Please give me some clues on this.
Virtually, any FS can be used on top of ScaleIO, but you don't want to use any FS that requires replication/segmentation of that within hosts at the FS level, because that, as you told, you compete with ScaleIO, but you could use a FS that has this functionalities, but you can choose not to use/not to enable them.
The point is that you are free to choose, since ScaleIO doesn't impose you any specific filesystem.
GlusterFS is pretty much the NFS protocol with automatic failover, and even if he is capable to spread-out files throughout different nodes, i see ScaleIO bringing more gains due the way how access/storage is striped among all the nodes of the cluster (my opinion). I would consider GFS as an option in a case where you want to export through NFS-Only and you don't require any local access to the share (not hyper-converged)...
Victor da Costa
I ve made up my mind, I will choose for now OCFS2 as I already used it on Oracle RAC and it worked very well indeed.Thanks for your replies!