Frequent policy does not result in periodic calls
Closed this issue · 3 comments
I'm trying to develop a plugin which calls the backend periodically for values. Therefore, I use the following policies: async, per_host, scorep_clock, frequent, object_id
I understood the docs in the way, that using frequent
should set Score-P to call stop
and get_all_values
every X ticks .
get_metric_gather_interval
returns 3 (for debugging only, bigger values result in the same behaviour) which I expect to make Score-P poll the backend quite often.
Nevertheless, get_metric_gather_interval
is called at initialization but stop
and get_all_values
are only called once at the end of the application.
Runtime of the traced application is about 3min and there are a lot of ENTER
and LEAVE
events uniformly distributed in this time.
Thanks,
Sebastian
The frequent
policy sets the delta_t
of Score-P metric plugins. However, this has a big disclaimer:
NOTE: This is only a lower limit for the time between two reads
If plugins need a periodic measurement, the usual way is to set up a measurement thread with the appropriate time interval between two consecutive measurements. See for example the x86_energy
plugin.
Ok. I thought this disclaimer states delta_t
is a lower limit even though an approximate time intervall.
I wonder that setting delta_t
to 3 ticks results in only one call in over 3 minutes. This seems strange to me.
Anyway, I will try to set up a measurement thread.
I'm going to close this, feel free to reopen if you need more information.