Hi Tianji, Yes, with some caveats. First, to set the heap size of the JVM, you can treat it just like any other Java program, e.g., see http://stackoverflow.com/questions/27681511/how-do-i-set-the-java-options-for-kafka <http://stackoverflow.com/questions/27681511/how-do-i-set-the-java-options-for-kafka> (the answer that starts with "You can pass java parameters from command line".
For a guideline on stream-specific setting see this thread: https://groups.google.com/forum/?pli=1#!searchin/confluent-platform/heap$20size$20streams%7Csort:relevance/confluent-platform/jazg8f-qn0M/hulJwk60AAAJ <https://groups.google.com/forum/?pli=1#!searchin/confluent-platform/heap$20size$20streams|sort:relevance/confluent-platform/jazg8f-qn0M/hulJwk60AAAJ>. One caveat is that RocksDb uses off-heap memory, but that can also be tuned. Thanks Eno > On 6 Apr 2017, at 14:23, Tianji Li <skyah...@gmail.com> wrote: > > If so, how to? > > Thanks > Tianji