(1) NIS server (nisserver), (2) file server (fserver), (3) cluster head server (queuemaster) (4) cluster compute nodes (node1,node2,...) (gateway=queuemaster)
fserver is a NIS client (to nisserver), queuemaster is a NIS slave server of nisserver, and the nodeX machines are NIS clients (to queuemaster). fserver exports several volumes to queuemaster and nodeX, and queuemaster exports several volumes to nodeX. All volumes are successfully mounted using NFSv4.
When I do a directory listing from nodeX of the volumes that queuemaster exports to nodeX, the user/group names are correctly listed, and using
ls -n shows the expected UID/GID values.
However, doing a list in the volumes that fserver exports to nodeX, the UID/GID and hence user/group information does not match the expected value, and instead gives nobody.nobody, yet doing a listing on the fserver itself does provide the correct UID/GID and user/group values.
I have tried restarting [rpcidmapd] after system boot and mounting, but to no avail.
What I do not understand is, given that fserver, queuemaster, and nodeX all use the same UID/GID information from nisserver, why are one set of volumes exported shown with the proper IDs while a second set are not?
In terms of network configuration, nissever, fsserver, and one port of queuemaster are on the same network, while the nodeX are a subnet which pass through a second network port of queuemaster to mount the volumes on fsserver.
ypcat group and
ypcat passwd retrieve the NIS information successfully.
If anyone can provide a hint, clue, or method of resolution, I'd be obliged.