The idea is that each mapper is only connecting to a single Kafka broker.
Each line in the input file specifies broker uri, topic, partition and
The hadoop consumer in contrib is probably a bit outdated. The one that
LinkedIn uses now can be found at https://github.com/linkedin/camus
On Tue, Jun 4, 2013 at 7:29 AM, Samir Madhavan <
[EMAIL PROTECTED]> wrote: