- UDP NFS mount of aliased IP addresses from a Solaris 10 NFS v4 server fail
- UDP NFS mounts of NFS v3 shares from the same server work fine
- Linux and Solaris clients work in default configuration when FreeBSD doesn't
- sysctl -w vfs.nfs.nfs_ip_paranoia=0 doesn't work even though mount_nfs -c does
- Two workarounds are possible:
  - mount_nfs -o tcp
  - mount_nfs -c

This problem was found in the following "Enterprise" setup:

- Server is a high availability cluster in which "service groups" will
  move as required between cluster nodes.
- "Service groups" have a static IP address associated with them (properly
  registered in DNS) which is the server name used by clients to NFS mount
  large amounts of storage.
- Server default IP address and all aliases are in the same subnet.
- Server software is: Solaris 10 with Symantec/Veritas HA 5.0 MP1 (aka
  Veritas Cluster Server).
- The Solaris 10 version is 11/06 with recommended and security patches
  of about 11 May 2007.
- In this arrangement we have several IP addresses on the one cluster
  node serving NFS and Samba.
- All Samba servers are bound specifically to the one "service group"
  IP address but there is only one NFS server per host such that it will
  respond on any of the IP addresses and will reply from the main hosts
  IP address instead of the "service group" address.

Since FreeBSD defaults to UDP mounts this makes FreeBSD fail straight
away (but all Solaris clients and Linux clients of several different
distros have worked fine to date).

Fix: 

Only two workarounds have been found:

- mount_nfs -o tcp server:/share/path /mount_point
- mount_nfs -c server:/share/path /mount_point

No fix is known.
How-To-Repeat: FreeBSD client:

# mount_nfs server:/share/path /mount_point

With "server" being an aliases IP address on a Solaris 10 box in default configuration (i.e. NFS version max is v4)