[SSI] Re: GFS NBD status?

[Brian J. W. <Bri...@co...>]

"Rouzer, Charles A (Chuck)" wrote:
> 
>         Does SSIC support read-only and read-write NFS mounts? 

It could probably be made to work. I think base Linux can run with a
read-write NFS root, so the SSI kernel should also be capable of this. A
drawback of NFS is that it has very loose coherency, meaning different
nodes will have a different idea of what's on the disk.

> Wondering
> what might be the best solution for a cheap test cluster.  I don't
> understand the point of using UML for a cluster except for development.  

The main purpose of UML is for development. It also allows new users to
easily set up a virtual cluster to play around with. I don't think it's
useful for production.

> Is
> my only solution to run the cluster using UML on 2+ different machines at
> this time?

Running a UML cluster across multiple machines is a future task. Right
now, Kitrick, John and I are just trying to get it to work on a single
machine.

For now, to run a real cluster without a shared drive, you can try to
find someone willing to get GNBD working with the latest OpenGFS, or
come up with a recipe for running SSI w/ a NFS root (assuming the
coherency problems are acceptable). You can also run the cluster without
a shared root, but this is problematic for migrating processes. In
particular, cluster_lilo won't work very well. Same goes for commands
that use cluster_lilo, such as addnode and chnode.

-- 
Brian Watson                | "Now I don't know, but I been told it's
Linux Kernel Developer      |  hard to run with the weight of gold,
Open SSI Clustering Project |  Other hand I heard it said, it's
Compaq Computer Corp        |  just as hard with the weight of lead."
Los Angeles, CA             |     -Robert Hunter, 1970

mailto:Bri...@co...
http://opensource.compaq.com/