Hi all, question...
Is it useful to create gluster volume on machines with one hard disc?
for example I have this: 1 physical machine with 7 nodes. Only 1 SSD per node.
Now I have shared storage on the master node that is shared across all other
nodes using NFS.
Is it handy to create distributed gluster volume across all nodes? Keeping in
mind there is only 1 SSD per node...
I was thinking about creating something like this:
Volume Name: sharedvolume
Type: Distribute
Number of Bricks: 8
Transport-type: tcp
Bricks:
Brick1: node0:/data/brick/shared
Brick2: node1:/data/brick/shared
Brick3: node2:/data/brick/shared
Brick4: node3:/data/brick/shared
Brick5: node4:/data/brick/shared
Brick6: node5:/data/brick/shared
Brick7: node6:/data/brick/shared
Brick8: node7:/data/brick/shared
And then share it on each node as:
nodeX:/sharedvolume /storage/shared glusterfs defaults,_netdev 0 0
A little more information on the set-up:
MASTER NODE (1x):
/dev/sda: 480.1 GB
/dev/mapper/centos_node0-root: 445.8 GB
/dev/mapper/centos_node0-swap: 33.8 GB
WORKER NODE (6x):
/dev/sda: 240.1 GB
/dev/mapper/centos_nodeX-root: 215.5 GB
/dev/mapper/centos_nodeX-swap: 24.0 GB
The root / needs a lot of space for /tmp. But the /storage/shared is also on
the root /.
It sounds dangerous to create another logical volume for the shared storage. If
the /tmp is full and the /storage/shared is not using a lot of space then this
extra LV is not handy.
David Comeyne
System Engineer
[http://www.applied-maths.com/pct/logo.gif]<http://www.applied-maths.com/default.htm>
_______________________________________________
Gluster-users mailing list
[email protected]
http://www.gluster.org/mailman/listinfo/gluster-users
