What I understand is the Kafka has following apis 1. Core broker APIs which runs the server. (Heart of kafka) 2. Producer API 3. Consumer API 4. Connect API 5. Streams API
All four use broker for different purpose. There should be some sort of compatibility matrix. Kafka server is something central to all applications and one where frequent upgrades are not possible in a production environment. So every release of other future APIs should have some backward compatibility with Kafka server. Like x.y.z.w server should atleast be compatible with x.y.**.** APIs. So any new release x.y.**.** should make sure that they are compatible oldest x.y.**.** release atleast. And perhaps some sort of compatibility matrix page should be made available. Just my 2 cents. Thanks Sachin On Sat, Dec 24, 2016 at 12:05 AM, Hans Jespersen <h...@confluent.io> wrote: > No. All Java clients (including Streams) need to be the same version (or > lower) as the brokers they connect to. > > -hans > > > > > > On Dec 23, 2016, at 1:03 AM, Sachin Mittal <sjmit...@gmail.com> wrote: > > > > Is Kafka streams 0.10.2.0-SNAPSHOT compatible with 0.10.0.1 broker. > > I was facing broker connect issue and it just did not work. > > > > Thanks > > Sachin > > > > > > On Fri, Dec 23, 2016 at 4:25 AM, <h...@confluent.io> wrote: > > > >> Kafka clients (currently) do not work against older Kafka > brokers/servers > >> so you have no other option but to upgrade to a 0.10.1.0 or higher Kafka > >> broker. > >> > >> -hans > >> > >>> On Dec 22, 2016, at 2:25 PM, Joanne Contact <joannenetw...@gmail.com> > >> wrote: > >>> > >>> Hello I have a program which requires 0.10.1.0 streams API. The jar is > >>> packaged by maven with all dependencies. I tried to consume a Kafka > >>> topic spit from a Kafka 9 cluster. > >>> > >>> It has such error: > >>> org.apache.kafka.common.protocol.types.SchemaException: Error reading > >>> field 'topic_metadata': Error reading array of size 1768180577, only > >>> 167 bytes available > >>> > >>> I wonder if there is any work around? > >>> > >>> Thanks, > >>> > >>> J > >> > >