Managing Systems and Workgroups: A Guide for HP-UX System Administrators

Setting Up and Administering an HP-UX NFS Diskless Cluster
NFS Diskless Questions and Answers
Chapter 10932
NFS Diskless Questions and Answers
This section answers some common questions about administering NFS
Diskless. It is a slightly condensed version of the “Questions and
Answers” section of the NFS Diskless Concepts and Administration
White Paper, which is supplied in its entirety as
/usr/share/doc/NFSD_Concepts_Admin.ps on most 10.x systems.
Cluster Configuration
Question: I have 100 clients and they all have the same kernel. Can the kernels be
linked to save disk space?
Answer: Yes. Client kernels (/export/tftpboot/ client /stand/vmunix) can be
hard linked with each other to save disk space.
It is also possible to hard link both the RAM file system (vmunix.fs) and
the LIF volume (uxbootlf). By default, the vmunix, vmunix.fs, and
uxbootlf files are hard linked with identical files used by other clients.
After the initial boot, the vmunix file is rebuilt by the client and the link
is broken.
Use only hard links, because:
• Symbolic links do not work for kernels.
This is because the /export/tftpboot/ client /stand directory is
mounted as /stand on a client system and any symbolic links within
this directory are resolved in the context of the client, not the server.
• HP does not support symbolic links for linking boot files.
Operations which modify the kernel or other boot files break any
existing links before writing a new boot file. This prevents a change
to one client’s boot file from affecting all clients that may have been
linked with that boot file. The best way to change all the clients’ boot
files is to change a single client and then re-establish the hard link.
Question: I have 100 clients, and they won’t all fit on the same disk.
Answer: You can spread the clients’ private directories and boot file directories
across multiple volumes. You can do this in one of two ways: