My only concern with this proposal is that it adds yet another way to specify configuration properties, making it a bit more difficult to track down how/whether a configuration property has been set. Configuring Kafka Connect is already too challenging, so we need to be very clear that this additional complexity is worth the price. IMO the KIP should explicitly address this.
Also, any reason why this KIP singles out the Connect distributed worker and doesn't address the standalone worker? And finally, the KIP does not clearly explain the command line behavior. It simply says: An additional command-line argument, --override key=value, will be accepted for ConnectDistributed. which makes it seem like only a single key-value pair can be specified. Clearly this is not the intention, but is `--override` used once and followed by multiple key-value pairs, or is `--override` required for every key-value pair? Does it need to follow the property file reference, or can the overrides precede or be interspersed around the property file reference? Does this happen to exactly match the broker command line behavior? The KIP should be very clear about the behavior and should fully specify all of these details. Best regards, Randall On Fri, Jun 15, 2018 at 11:14 AM, Jakub Scholz <ja...@scholz.cz> wrote: > I think this makes perfect sense. Thanks for opening this KIP. > > Thanks & Regards > Jakub > > On Fri, Jun 15, 2018 at 2:10 AM Kevin Lafferty <kevin.laffe...@gmail.com> > wrote: > > > Hi all, > > > > I created KIP-316, and I would like to initiate discussion. > > > > The KIP is here: > > > > https://cwiki.apache.org/confluence/display/KAFKA/KIP- > 316%3A+Command-line+overrides+for+ConnectDistributed+worker+properties > > > > Thanks, > > Kevin > > >