Hi Marko,
----- Original Message -----
> Hi,
>
> On 2016-03-07 23:44, Nathan Scott wrote:
> > ----- Original Message -----
> >> [...]
> >> It would seem that the limitation still exists for PMDAs, see the below
> >> test PMDA - it works as expected if the innermost loop is removed from
> >> generate_metrics() (676 metrics added) but if the innermost loop is in
> >> place, then only 1024 metrics are added even though there are over 17k
> >> metrics in the hash:
> >> [...]
> >> $pmda->add_metric(pmda_pmid(0, $i++),
> >
> > You're not using the cluster space (the 0 above) - only item ($i above),
> > which is limiting the number of metrics you can have artificially here.
>
> Confirmed, by putting max 1024 metrics per cluster space all the ~17k
> metrics can be added for the PMDA.
>
> This makes me wonder shouldn't the user get some sort of warning about
> the situation, now data is silently discarded in a non-obvious manner?
>
Yep, for sure, more checks would help with debugging this kind of issue.
The current code comes from a world of PMDAs where metric IDs are mainly
statically defined via metric tables, so the setup macros are simple and
branch-less - but the python/perl APIs could certainly improve on that.
cheers.
--
Nathan
|