Geoff Hardin ghardin@dalsemi.com writes [...]
By and large, if we have individual NFS clients attached during a reboot, they can the 30 seconds it takes for the filer to reboot. We have had isolated instances where older SunOS systems (and maybe our Solaris 2.6 FlexLM license servers) have a problem with automounted NFS mounts, so those are usually shutdown in the event we have some planned downtime. Other than that, most people just blame it on the network and keep on going.
We have had a problem with the automounter even in recent versions of Solaris (including one instance with Solaris 9 so far). It seems to be triggered by the automounter deciding to try unmounting an NFS filing system while the server is unresponsive. The result is that that all accesses to the mounted filing system stall indefinitely (but interruptibly) even after the server is back.
This is low-frequency for any particular server reboot, but still sufficiently common that we try to schedule things so that we can follow up the server reboot by a client reboot if it strikes.
Chris Thompson Email: cet1@cam.ac.uk