SIAC has found a bug with the 5.0.1 OS of DataONTAP. There appears to be no time-to-live on the cached IP addresses of hosts. Changes in the IP addresses of hosts in DNS or NIS are not recognized until the NetApp is either rebooted or "exportfs -a" is executed. To force re-reading of DNS and NIS data, we have set up cronjobs to reboot our 6 NetApp 520's weekly.
Does anyone know if this is also a problem w/ DataONTAP 4.x?
It's also a problem with SunOS 4.x, and probably 5.x, as well, and probably most if not all other UNIX NFS servers.
The problem isn't that the NIS or DNS code (or, for that matter, the code that reads the filer's "/etc/hosts" file) cache the result of hostname lookups.
The problem is that "exportfs" translates the host names in the "root=" and "rw=" lists to IP addresses, and the IP addresses are what's stored in the exports list; SunOS (and probably all UNIX systems whose NFS servers are based on the Sun NFS releases, and possibly those that aren't) does the same.
"exportfs -a" rereads the "/etc/exports" file and re-translates the IP addresses. (Rebooting also does that, but it's overkill....)