On Thu, Apr 28, 2016 at 1:03 PM, Jeff Schroeder <jeffschroe...@computer.org> wrote:
> On Monday, April 25, 2016, John Sirois <john.sir...@gmail.com> wrote: > > > On Mon, Apr 25, 2016 at 8:15 PM, John Sirois <john.sir...@gmail.com > > <javascript:;>> wrote: > > > > > > > > > > > On Mon, Apr 25, 2016 at 7:58 PM, Bill Farner <wfar...@apache.org > > <javascript:;>> wrote: > > > > > >> I don't have a firm example in mind, I just don't think the approach > > >> recommended by Zameer is optimal. It's only marginally better than > > >> forking > > >> (and probably requires you to fork for the sake of sanity). > > >> > > >> Thinking out loud, it would be nice if Kafka had a JMS > interface/bridge, > > >> as > > >> it would allow us to add support for a bunch of backends with one > > >> implementation. Unfortunately that does not appear to be the case. > > >> > > > > > > That said - supporting a JMS interface doesn't sound too bad on the > > aurora > > > end. > > > Motivated consumers could write a bridge presumably if they insist on > > > using Kafka. > > > > > > > Or maybe even just a webhook API. You configure aurora with an HTTP > > endpoint that must conform to a given api and Aurora tries, best effort > > only, to post events to the endpoint, perhaps on a streaming connection. > > This scales well for Aurora. > > > > Or Server Send Events, an often overlooked but really nice way of having a > server send clients events in realtime over http. I *think* this is the way > marathon sends events to marathon-lb (haven't looked though) > The reason I leaned towards 1 webhook url configured on the aurora scheduler command line was paranoia over load mainly. This assures aurora it will only have to service 1 best-effort streaming client instead of an arbitrary number, and it assures aurora that only operators can choose to enable this instead of random joe dev #9165 out of 12000 at some huge company. I guess that concern is also easily assuaged though via admin-only api for example. > > > > > > -- > Text by Jeff, typos by iPhone >
