Steve, I think exposing perhaps "instance id" in agent api should be good enough. I don't think people need full override policy for configuration of every instance.
Moreover, do you preserve locations for each instance? For a service like kafka, it writes data to local disk. You can not really switch broker id in between instances after relaunch. Thanks! On Thu, Jan 29, 2015 at 2:44 AM, Steve Loughran <ste...@hortonworks.com> wrote: > I see. There's configuration per component type, but not per instance of > that type. > > That's an interesting idea...I'll have to think about a way that could be > made to work > > ________________________________________ > From: 杨浩 <yangha...@gmail.com> > Sent: 28 January 2015 00:45 > To: dev@slider.incubator.apache.org > Subject: Re: Component instance level configuration > > do you mean each instance of the. same component will have their own > configuation? Only port can be different for instance of a component > nowadays. > > 2015-01-28 5:57 GMT+08:00 hsy...@gmail.com <hsy...@gmail.com>: > > > Hi, > > > > I'm just wondering if there is a way to support component instance level > > configuration. For example, in kafka-on-yarn, kafka broker is a > component, > > but each broker instance needs a unique id. It would be better to expose > > the instance id, so that different instance would override some of the > > configurations accordingly > > > > Best, > > Siyuan > > >