$ ./bin/kafka-console-consumer \
--bootstrap-server localhost:9092 \
--from-beginning \
--property print.key=true \
--topic connect-offsets
["jdbc_source_mysql_foobar_01",{"table":"foobar"}] {"timestamp_nanos":0,"timestamp":1501796305000}
["jdbc_source_mysql_smt_foobar_02",{"table":"wibble"}] {"timestamp_nanos":0,"timestamp":1501800342000}
["jdbc_source_mysql_smt_foobar_02",{"table":"wibble"}] {"timestamp_nanos":0,"timestamp":1501800707000}
Not all connectors will use this. For example, the HDFS connector uses HDFS itself to track offsets as part of the exactly-once delivery mechanics.
The key is the connector name and additional elements (e.g. for the JDBC connector the table). The value is the offset being tracked.
This is only for Kafka Connect in Distributed mode. When run standalone a local file is used.