[gpfsug-discuss] Filesystem access issues via CES NFS

Jonathan Buzzard jonathan.buzzard at strath.ac.uk
Wed Oct 23 12:20:18 BST 2019


On Wed, 2019-10-23 at 10:14 +0000, Simon Thompson wrote:
> From our experience, you can generally upgrade the GPFS code node by
> node, but the SMB code has to be identical on all nodes. So that's
> basically a do it one day and cross your fingers it doesn't break
> moment... but it is disruptive as well as you have to stop SMB to do
> the upgrade. I think there is a long standing RFE open on this about
> non disruptive SMB upgrades...
> 

My understanding is that the issue is the ctdb database suffers from
basically being a "memory dump", so a change in the code can effect the
database so all the nodes have to be the same. It's the same issue that
historically plagued Microsoft Office file formats.

Though of course you might get lucky and it just works. I have in the
past in the days of role your own because there was no such thing as
IBM provided Samba for GPFS done exactly that on several occasions.
There was not warnings not to at the time...

If you want to do testing before deployment a test cluster is the way
forward.

JAB.

-- 
Jonathan A. Buzzard                         Tel: +44141-5483420
HPC System Administrator, ARCHIE-WeSt.
University of Strathclyde, John Anderson Building, Glasgow. G4 0NG





More information about the gpfsug-discuss mailing list