GlusterFS + DB

[Gerry Reno <greno@xxxxxxxxxxx>]

 Lately I've just been lurking and reading all the posts.  Amazing 
progress by all the GlusterFS developers...  Looks like 1.3 is getting 
pretty stable and I am anxiously awaiting GlusterFS 1.4 and self-heal 
(as is everyone else here I'm sure).  That is where we can take 
GlusterFS out of our labs and onto the track so to speak.  Now I'm 
beginning to think about some of the issues of running a 
shared-filesystem database cluster.  Existing methods of doing this have 
usually involved SANs and FC and GFS and similar and insuring that 
writes to the block device were updated to the cluster nodes in near 
real time.  So is GlusterFS suited to being used in a shared-filesystem 
database cluster scenario?  Can the performance be high enough that the 
database cluster nodes will not see different/corrupted data?  How can 
data integrity be addressed?  When 1.4 is somewhat complete I am 
thinking about setting up a simple AFR cluster and running a simple 
unpartitioned database over it just to see if this is possible.

Regards,
Gerry