Hi,
we're using nfs option soft for all mount off of a filer. My goal is to tune NFS parameters so that I can do a quick filer reboot without currently running jobs giving NFS errors.
First UDP. On Solaris, default value of retrans is 5 and default value of timeo is 11, and the way I understand it that I have 5*11/10 = 5.5sec of allowable NFS unavailability before getting fatal NFS errors - not good enough for a filer reboot. So I can set, for example retrans=50,timeo=60, or retrans=15,timeo=200 etc., in order to extend the *grace* period to 5min. My guess is higher timeo/lower retrans is better than vice versa.
Now TCP. On Solaris, default value of timeo is 600 (1min), and retrans is N/A, so does that mean that if my only option is to set timeo=3000 to achieve the same goal?
Any comments would be appreciated.
Thanks
Igor