I have a main storage cluster running here (GPFS 3.4.07 on Linux). Call this Cluster A.
I want to create another (virtual) cluster - same OS+GPFS versions - which remote mounts my first cluster. Call this Cluster B.
Cluster B is for researchers who want high speed access to Cluster A data, but they might do "crashy" experiments. Remote mount seems like a good idea so as not to impact Cluster A when nodes in B crash.
However, I have some security / ACL concerns. I know I can "root squash" Cluster B when it mounts Cluster A. I don't trust these researchers not to poke around and break stuff. They want root access to their development machines. Fine, we routinely give this.
So, when it comes to UID / GID mapping from cluster B to Cluster A, what happens? Can a user with UID=10000 in Cluster B access data on Cluster A which is owned by UID=10000 there? I remember reading a very long time ago about GPFS GUNs (Globally Unique Names) but can't find the doc now - am I thinking in the right area? How does one manage these mappings if they exist?
Notice: We have upgraded developerWorks Community to the latest version of IBM Connections. For more information, read our upgrade FAQ.