I think we probably can drop this completely. The new selector should have provided enough metrics to describe the IO thread status. And it is async anyway, so ³being sent² does not look really helpful.
Jiangjie (Becket) Qin On 5/14/15, 1:06 PM, "Todd Palino" <tpal...@gmail.com> wrote: >I don't believe we're using the ResponsesBeingSent information at all. I >know we use NetworkProcessorAvgIdlePercent to keep track of the >utilization >of the pool. > >-Todd > > >On Thu, May 14, 2015 at 11:36 AM, Joel Koshy <jjkosh...@gmail.com> wrote: > >> I'm also not sure how useful it is, but there is some discussion on it >> here: >> >> https://issues.apache.org/jira/browse/KAFKA-1597 >> >> On Thu, May 14, 2015 at 09:26:02PM +0300, Gwen Shapira wrote: >> > Hi, >> > >> > As part of KAFKA-1928, we need to consolidate existing SocketServer >> metrics >> > into Selector metrics. >> > >> > The two metrics are: >> > 1. NetworkProcessorAvgIdlePercent >> > 2. ResponsesBeingSent >> > >> > NetworkProcessorAvgIdlePercent is currently implemented as Meter, >>which >> > seems to translate nicely into KafkaMetrics Rate. >> > >> > ResponsesBeingSent is a gauge, which doesn't have a direct >>KafkaMetrics >> > equivalent, but should be easy to add. >> > >> > However, I'm considering dropping ResponsesBeingSent gauge completely >>- I >> > find it useless. >> > Selector currently has Rates for requests-sent, responses-sent, >> bytes-sent, >> > bytes-received. >> > >> > Anyone finds ResponsesBeingSent useful and wants to make a passionate >> plea >> > to keep it around? >> > >> > (I CCed the people I think will care, in case they are not watching >>the >> > list) >> > >> > Gwen >> >>
