I wasn't aware of this. No objections to changing it. In fact, I actively encourage it.
-Ivan On 14 November 2014 01:56, Sijie Guo <guosi...@gmail.com> wrote: > Hi guys: > > I am revisiting the auto-recovery code. There were tons of synchronous > calls in asynchronous callbacks (zookeeper/bookkeeper callbacks), which > potentially will block the zookeeper/bookkeeper client worker threads. > Specifically there isn't asynchronous method for > LedgerUnderreplicationManager interface. This is very bad practice. > > Ivan/Rakesh, are you guys aware of this? I am going to make a change to > make it asynchronous. Any objections? > > - Sijie