Hi Cheng, On the KIP page, it lists this KIP as "draft." It seems like "under discussion" is appropriate here, right?
> Currently, the initial socket connection timeout is depending on Linux kernel > setting > tcp_syn_retries. The timeout value is 2 ^ (tcp_sync_retries + 1) - 1 seconds. > For the > reasons below, we want to control the client-side socket timeout directly > using > configuration files Linux is just one example of an OS that Kafka could run on, right? You could also be running on MacOS, for example. > I'm proposing to do a lazy socket connection time out. That is, we only check > if > we need to timeout a socket when we consider the corresponding node as a > candidate in the node provider. The NodeProvider is an AdminClient abstraction, right? Why wouldn't we implement a connection setup timeout for all clients, not just AdminClient? best, Colin On Mon, May 4, 2020, at 13:18, Colin McCabe wrote: > Hmm. A big part of the reason behind the KIP is that the default > connection timeout behavior of the OS doesn't work for Kafka, right? > For example, on Linux, if we wait 127 seconds for a connection attempt > to time out, we won't get a chance to make another attempt in most > cases. So I think it makes sense to set a shorter default. > > best, > Colin > > > On Mon, May 4, 2020, at 09:44, Jose Garcia Sancio wrote: > > Thanks for the KIP Cheng, > > > > > The default value will be 10 seconds. > > > > I think we should make the default the current behavior. Meaning the > > default should leverage the default connect timeout from the operating > > system. > > > > > Proposed Changes > > > > I don't fully understand this section. It seems like it is mainly > > focused on the problem with the current implementation. Can you > > explain how the proposed changes solve the problem? > > > > Thanks. > > > > > > -- > > -Jose > > >
