[Gluster-users] Re. Any update pls : Req details for "config as like striping with fail over" for many servers

Dan Bretherton d.a.bretherton at reading.ac.uk
Sat Apr 2 09:59:51 UTC 2011

> I see that
> your backend sizes are different... Its preferred  to keep them uniform.

> try to keep the backend uniform to avoid any possible issues which may
> arise later.

Hello All-
These comments worry me because I have unequal sized bricks in my 
replicated-distributed volumes, although every replicated brick is the 
same size as its mirror. I am not using striping.  What are the 
"possible issues that might arise" with with setup, or does this advice 
only apply to striped volumes?

A comment later on in this thread suggested the problems with 
non-uniform brick sizes were performance related.  If that is the case 
how much could I improve performance by migrating to uniform brick sizes?

I would prefer not to do this because my current setup gives me a lot of 
flexibility.  When a volume needs expanding I usually increase the size 
of the replicated bricks on a pair of servers that have unallocated 
space, or if new servers are required I create new bricks the size of 
the extra space the users require.  In other words, I don't extend the 
volume by a particular amount just because the other bricks happen to be 
that size.

Any advice would be much appreciated.

More information about the Gluster-users mailing list