20/10/2020 17:02, Stephen Hemminger: > Thomas Monjalon <tho...@monjalon.net> wrote: > > 05/09/2020 00:31, Stephen Hemminger: > > > If using lots of queues and ports, and having per port or per queue > > > metrics it is easy to exceed the upper bound of the metric library. > > > Move the limit into rte_config where user can change it. > > > > > > Ideally, there would be no upper bound and a dynamic structure > > > such as red-black tree or hash table would be used for these. > > > > > > Signed-off-by: Stephen Hemminger <step...@networkplumber.org> > > > --- > > > --- a/config/rte_config.h > > > +++ b/config/rte_config.h > > > +/* rte_metrics defines */ > > > +#define RTE_METRICS_MAX_METRICS 256 > > > > Not sure we want to go in the direction of adding such tuning > > in rte_config.h. > > The only other option is to rewrite rte_metrics to support a > dynamic data structure!
Yes it seems a better solution. How rte_metrics will fit in the picture of the new telemetry? Maybe we will deprecate rte_metrics in future?
