sounds great. On Tue, Oct 16, 2018 at 2:20 PM yuhang xiu <carry...@gmail.com> wrote:
> Hi, Ian! > > Not yet, I am starting this work. When I am done, I will send an email > immediately. > > Ian Luo <ian....@gmail.com> 于2018年10月16日周二 下午2:15写道: > > > good proposal, we should definitely switch to HashWheelTimer to make the > > timer execute in more fair/timely manner. > > > > Is there any pull request ready for review? > > > > Thanks, > > -Ian. > > > > On Thu, Oct 11, 2018 at 10:41 AM yuhang xiu <carry...@gmail.com> wrote: > > > > > Hi, community > > > > > > I recently saw an issue[1] from @chickenlj about the frequency of > > heartbeat > > > detection. Due to the very low frequency of our heartbeat detection, > some > > > idle channels may not send heartbeats in time. > > > > > > I will start working on this issue in the near future, as follows: > > > 1. Use hashwheeltimer to increase the granularity of the check > > > 2. Currently all channels of a client use the same heartbeattask. I > will > > > change to use each heartbeattask for each channel. > > > 3. In addition, the current reconnect operation is included in the > > > heartbeattask. In fact, this is not necessary. I will split the > heartbeat > > > and reconnect operations into two tasks for more precise control. > > > > > > What do u think about it? > > > > > > [1] https://github.com/apache/incubator-dubbo/issues/2537 > > > > > >