We do have mesos based infrastructure to run our java-based microservices
and we'd like to use it for deploying connectors as well (with benefits of
reusing deployment specific knowledge we already have, isolating the load
and in general pretty much the same reasons Kafka Streams was designed to
be a library instead of server). We programmatically start connect
framework from within and it works just fine, with one exception - we have
to replicate 3 connect framework internal topics (statuses, configs,
offsets) for each connector we run this way.
I wonder if I miss something here and/or if there are any plans for making
connect framework more friendly towards this type of deployments?