[gpfsug-discuss] CES behind DNS RR and 16 group limitation?

[Dietrich, Stefan]

Hello, 

we will soon start to deploy CES in our clusters, however two questions popped up. 

- According to the "CES NFS Support" in the "Implementing Cluster Export Services" documentation, DNS round-robin might lead to corrupted data with NFSv3: 
If a DNS Round Robin (RR) entry name is used to mount an NFSv3 export, data corruption and data unavailability might occur. The lock manager on the GPFS file system is not clustered-system-aware. 

The documentation does not state anything about NFSv4, so this restriction does not apply? 
Has somebody already experience with NFS and SMB mounts/exports behind a DNS RR entry? 

- For NFSv3 there is the known 16 supplementary group limitation. 
The CES option MANAGE_GIDS lifts this limitation and group lookup is performed on the protocl node itself. 
However, the NFS version is not mentioned in the docs.

Would this work for NFSv4 with secType=sys as well or is this limited to NFSv3?
With NFSv4 and secType=krb the 16 group limit does not apply, but I can think of some use-cases where the ticket handling might be problematic. 

Regards, 
Stefan

--
------------------------------------------------------------------------
Stefan Dietrich            Deutsches Elektronen-Synchrotron (IT-Systems)
                        Ein Forschungszentrum der Helmholtz-Gemeinschaft
                                                            Notkestr. 85
phone:  +49-40-8998-4696                                   22607 Hamburg
e-mail: stefan.dietrich at desy.de                                  Germany
------------------------------------------------------------------------