Hi,
I am using Storm 1.1.0 ,storm kafka client version 1.1.1 and Kafka server
is 0.10.1.1.

Kakfa spout polling strategy used is UNCOMMITTED_EARLIEST.

Message flow is like below and its a normal topology

KafkaSpout --> AvroDeserializerBolt-->DataBaseInsertBolt.

If the message fails avro deserialization , i am moving the message to a
error queue and acknowledging from the avro bolt . This message is not
emitted to database bolt .

But its observed that after i restart topology , offset for the topic is
going back to old offset.

Will Kafka commit the offset, only if the message is acked from all bolts ?

Is the offset going back to previous value is beacuse of this ..

Thanks
Pradeep

Reply via email to