[Gluster-devel] More questions

[Gerry Reno]

>  Remember early when I asked about being able to have all bricks 
> update and read from any client using AFR?  Well of course this seems 
> to work fine in my small limited tests.  Would this work well enough 
> that you could place a database on the bricks and have it be 
> consistent no matter which brick was being accessed by clients?  If 
> so, then I would still have to work out details of how to snapshot 
> backups (probably will use LVM) and how to control and administer the 
> db but it seems like it could be an alternative to other database 
> replication methods.  Right now these require that you find a way to 
> split the reads and writes between masters and slaves.  I would like 
> to get away from having to deal with that requirement and perhaps 
> glusterfs could help here.
>
I've been researching this today and it looks like MySQL may have 
improved their NDB Cluster solution and removed some of the limitations 
that it had.  It's still not perfect but now you can have data on disk 
rather than in memory.  Not indexes though.  So for indexed columns you 
have to tell it to use a different storage.  Anyway, what I'm looking 
for is a completely transparent cluster solution for the database.  I 
guess we're not there yet.

Gerry