So you are publishing to kafka brokers running at client networks? And they are consuming locally to brokers on their networks?
Why not publish to brokers running locally to you and have them consume it over the internet (even through an API possibly). Yes you have to now operate the broker but then you don't have to deal with 3 different clients being down because you have 0% control over what you are pushing to. Hard to scale that way. My $0.0244532. to-do what I think you are suggesting would require you to basically have to provide support & maintenance for Kafka and every permutation that everyone has to deal with, like as a vendor. /******************************************* Joe Stein Founder, Principal Consultant Big Data Open Source Security LLC http://www.stealth.ly Twitter: @allthingshadoop <http://www.twitter.com/allthingshadoop> ********************************************/ On Thu, Dec 26, 2013 at 5:10 AM, Amir Gershman <[email protected]>wrote: > Hi, > > I am looking into using Kafka for production to store notifications sent by > one of our services. The service clients will then be able to either replay > lost notifications while they were down or simply consume the notifications > directly from Kafka (and not from the service) at their own pace (for > example at low system load times). > Our software is deployed at each customer's local datacenter, so the > operations and maintenance are done by our clients. Our clients typically > operate windows servers, so we wish to deploy Kafka on windows. > I imagine something like JMS persistent replay is a common use scenario. > Does someone have some useful suggestions on this matter, or things to > watch out in running Kafka on windows. > > Thanks, > Amir. >
