2011-07-08 07:47 AM
I have a problem with perfstat.sh taking a long time (~5-10 minutes) to run the prestats. The poststats take almost the same amount of time, maybe a couple minutes shorter. The following is the invocation used:
perfstat.sh -F -S -f <filer> -l <user> -t 1 -i 1 > <path to directory>/perfstat.out
I have password-less SSH set up on the filer for my user.
I am trying to capture the data on the filer while a simulated load is put on it. The load is generated via filersio and runs for 5 minutes. The problem is that the simulated load is over before perfstat finishes the prestats. This results in the perfstat output data to be null. One perfstat iteration takes about 2-3 times as long as the load, which is not beneficial.
When looking at the perfstat in real time, the commands issued take about ~1-5 seconds, with 33 seconds being the longest. The following links are sample perfstats of the problem:
What are the causes of perfstat.sh taking so long to run? And how do you fix it?
Thanks for the help,
2011-07-08 09:08 AM
I've noticed some improvement in using different cycle servers. By switching to an RTP cycle server, perfstat.sh doesn't run as slow. Before I was using a cycle server out of NANE.