I assume you meant Kafka offset - https://kafka.apache.org/documentation/#intro_topics
Currently I don't think this is possible due to two reasons. (1) Currently Kafka source can either read from a given topic or a set of topic partitions, but not from a given offset - https://github.com/apache/beam/blob/master/sdks/java/io/kafka/src/main/java/org/apache/beam/sdk/io/kafka/KafkaIO.java#L523 (2) Currently source has to be at the top of the pipeline graph to operate. For example, you cannot initiate a Kafka source from a topic/partition read from a database. (2) should be possible when we have our next generation source framework, SplittableDoFn. May be when we have that we can consider adding (1) as well, if there are good justifications for that. I think the policy regarding the offset to start reading from is configured in the Kafka cluster and specifying a specific offset will not work if the corresponding messages have been purged by the Kafka cluster, so I'm not sure how useful adding support for reading from a given offset will be. Thanks, Cham
