On Mon, May 16, 2016 at 8:35 PM, Colin McCabe <[email protected]> wrote:
> Hi Lewis, > > Thanks for trying this out. The htraced buffer gets flushed after a > configurable amount of time. It's also flushed if the SpanReceiver gets > shut down (like in a short lived command line program.). So I don't > think we should change the buffer size default. Maybe we could document > this better, and/or change the buffer flush time default. > > The latter two ideas sound good to me. It can be confusing when you don't see spans when you expect too. It can take a while to figure the why not. St.Ack > Best, > Colin > > On Mon, May 16, 2016, at 19:34, Lewis John Mcgibbney wrote: > > Hi Folks, > > For test and development and also for new comers coming to HTrace I am > > not > > sure if it is realistic for such a high default buffer to be set for > > Htraced. Right now is appears that the default value is 16,777,216 bytes > > which is pretty large if one wishes to get a feel for the HTrace > > framework > > generally and HTraced specifically. > > Although I am easily able to work programmatically and hence change this > > default buffer value to anything I want. Everyone else may not be that > > way > > inclined and may instead just wish to, fr example, run the Docker image > > and > > see what HTrace actually does. > > Does anyone else have an opinion on this? My perspective from being > > @ApacheCon last week and from doing my HTrac'ing in Nutch presentation is > > that people just want to see what HTrac'ing looks like. They may not have > > an application which writes 16MB of spans to a buffer out of the box. > > Thanks > > Lewis > > > > -- > > *Lewis* >
