I have slow running Oracle RAC database on NFS volume.
I think I have zeroed it but some clarification.
I am using nfsiostat in RHEL OS host to monitor the NFS share.
I have just kicked off the script to monitor the nfsiostat stats and can see massive difference in avg RTT (ms) and avg exe (ms)
RTT is under 10 msec however avg exe is between 100 to 200 msec.
I got to understand below
RTT = NFS Server + Network time
Avg exec = time between kernel call by Oracle process to kernel call close
so avg exe is end to end time RTT inclusive.
is above understanding correct ?
if so by seeing the recent stats which I have , I can certinly rule out NFS server and Network here
and the prime suspect is OS kernel === how effieicently kernel is handling the requests.
I have database performance statistics as well , which points to it to be a issue with how kernel / OS is handling these requests .
open_files is one of the key suspect and bug in Oracle cluster may be adding.
your expert inputs on above please.