[Gluster-devel] Re :Confusion GlusterFS over GFS(Global File system)

Gordan Bobic gordan at bobich.net
Sun Nov 9 12:06:53 UTC 2008


Yang Ye wrote:

> GlusterFS exports a file system. Once mounted, it works as like local 
> hard disk, like NFS. So there is no relationship between MyISAM and 
> GlusterFS. And you can definitely mount GlusterFS volume to 
> /var/lib/mysql, though I recommened to mount it somewhere else and 
> configure mysql to that directory. This answers (4) and (7).

I wouldn't want to try concurrent access from multiple nodes to a MySQL 
DB. I really don't think that would work without trashing the data.

> For (1), (2) and (3), better try yourself. To me, GFS configuration is 
> not something I would like to try. It's just too complex, IMHO.

GFS isn't _THAT_ hard to configure when you consider that GFS doesn't 
come on it's own - you have to configure the clustering subsystem to 
guarantee that the file system stays consistent, which also requires 
external fencing. It depends on whether your application requires extra 
consistency that GFS guarantees. In GFS, concurrent write locking is 
handled implicitly and you cannot have concurrent write locks from 
several nodes. GlusterFS doesn't even try to pretend that this is a 
reasonable use case (locking is handled using advisory POSIX locks). But 
the relaxation of restrictions does give a file system that has better 
performance in it's intended field of application. The applications of 
GFS and GlusterFS are sufficiently divergent that there is seldom 
justification to consider them equivalent.

Gordan





More information about the Gluster-devel mailing list