[Gluster-devel] general questions

[Corin Langosch]

Hi!

I'm pretty new to clustered storage but I really need it due to high 
availabilty, amount of space needed and performance reasons for my 
disk-less clients.

As for now I have a big server running a RAID5 exported as a block 
device using iSCSI. I use ocfs2 to mount the device on the clients. So 
far it works, but I'm not happy with it as I still have a single point 
of failure, only all my clients can run without a disk and access the 
same data. Using nfs is not an option as it's way to slow.

I just played a little with glusterfs (1.3.7) and I'm simply amazed by 
it's easy of use. As this filesystem seems pretty new (didn't find a 
history on the site when it started, but found only very little 
information in wikipedia, google, no official debian packages, etc..) I 
wonder if it's ready to be used in production environments.

My goal is to have 2-3 big storage servers for now. The storage of these 
servers should exported unified with automatic replication. Currently I 
have 25 clients, but I'll need to increase to 100-200 clients soon. The 
clients to not generate a lof of traffic, so 2-3 servers should be ok 
for now.

The my remaining questions are:
- is glusterfs the right choice for me? (if not, what is? *g*)
- is the current code (1.3.7) stable enough? no dataloss expected?
- if data is stored on all three nodes using automatic replication, can 
I simply drop a server if it's damaged, replace the disk etc., put it 
back online and it'll automatically be reused?

Corin