odd perfmon metrics after change of granularity

[cyril....@gmail.com]

Hi, I use the option --interval on my agent.

I set the interval to 30s. I note that the Disk I/O writes metrics are weird now.

It seems that the value is equal to the number of writes during the last 30 seconds, not the average writes per seconds during the last 30 second.

For example, if there are 10 writes per seconds on average on the disk, the graph will show 300 writes.
Is it a bug ? or there is a way to change the configuration to obtain 10 write/s.

Thanks

[Vincent Daburon]

Hi,

#Perfmon metrics collection interval in milliseconds
jmeterPlugin.perfmon.interval=4000

Every 4 seconds, Perfmon Listener demand current values to the perfmon agent.
It's polling not cumulated values.

Regards.
Vincent D.

[Andrey Pokhilko]

You are right, it's not "per second". And documentation does not say it.

However, I agree with you, maybe it is more sense to always translate some of the metrics into per-second values.

--

Andrey

пятница, 6 февраля 2015 г., 4:47:15 UTC-8 пользователь cyril....@gmail.com написал:

[cyril....@gmail.com]

thanks, 

Just to be clear, the jmeterPlugin.perfmon.interval property changes the sample rate.

So, if there is lot of variation in the metric, with big interval, the information can be loss.

The interval option of server agent seems to agregate the value beetween two polling period, so the interval can be big.

am I right ?