NFS after takeover

[Peter Silva]

Tim Burke wrote:
> 
> Sorry to beat on this issue but.......... if there's no room in the NFS
> file handle to store both the 64bit OID and the major/minor number of
> the device, then how is it handled even in a non-failover scenario?
> Presumably ReiserFs would map its own OID to the associated major/minor
> number.  If the major/minor numbers referring to the shared storage were
> the same on all cluster members then would the failover work?
> 

NFS servers makes an arbitrary "fsid" (usually) based on the device used.
How the hardware makes an fsid is a (server) platform specific issue.
It is the fsid that is assigned by the server on export.
The client receives this on mount and passes it to the server
with every io request for that file system.

I don't think one can say much about major and minor numbers,
they used to be catenated into the "rdev" field, but that
was way back when both quantities were 16 bit.  I do not
see how major/minor devices are available to an NFS client
at all anymore.  

-- 
mailto:Peter.Silva@ec.gc.ca  -  Supercomputing, Environment Canada