We managed to workaround the issue by rebuidling a package and using a custom Hortonworks repository with Ambari. Everything's just fine :-)
*Benjamin Delespierre **| Developper* [email protected] Skype : benjamin.delespierre.75 Hangout : [email protected] 21 rue de Châtillon - 75014 Paris - France Facebook <http://www.facebook.com/AdServerPub> | Twitter <http://twitter.com/AdServerPub> | AdServerPub.com <http://www.adserverpub.com/> 2015-07-28 4:48 GMT+02:00 Artem Ervits <[email protected]>: > I faced with similar issue. I was using Storm and Hive. Take a look at > Hive streaming injest API, it uses the same technique as flume Hive sink. > On Jul 21, 2015 11:13 AM, "Benjamin Delespierre" < > [email protected]> wrote: > >> Hello, >> >> We're using Kafka for event retention and would like to stream the data >> directly into Hive using Flume. The problem is that only Flume 1.6 can acts >> as a Kafka client for now and it's not yet in the Hortonworks stack >> (currently HDP 2.3 still use the 1.5). >> >> How can I get the proper flume version installed ? >> >> - Should I create a custom repository with my package in it (can't get my >> hands on a .rpm or .deb for this version by the way) ? >> - Should I create a custom service ? >> - Should I modify the existing service in order to change the >> installation behavior ? >> - Am I completely wrong ? >> >> Any help is welcome ;-) >> >> Best regards. >> -- Benjamin >> >>
