setting Kafka offset commit count

@rbeerak

Manually committing offsets is currently not supported in the KafakConnector. One thing you could try is the advanced options in the KafkaConnector. You could experiment with the auto.commit.interval.ms parameter to overwrite the default 5000 ms. If you decrease the interval the consumed messages will be committed earlier. Since you know the number of messages in your batch you might be able to come up with an interval that commits the messages right after they are consumed.

@gerhardatsafe I have used maximum value for auto.commit.interval.ms. However, I still do not have control over committing kafka offset.

Here is my job flow/configuration:

1. Kafka connector (acts as consumer).

Configured to run in -

"BATCH" mode

"BATCH SIZE":2000 (# of messages per batch)

"EARLIEST" for starting offset.

{"auto.commit.interval.ms":500000}

2. JSON parsing

3. Load parsed tags into postgresql with commit count:2000

Issue:

When there is some issue with table load (probably data conversion) during job execution, the job obviously fails with error and kafka offset gets committed. This means when I rerun the job fixing table, data consumption is happening from next set of messages and messages previous failed batch are lost.

I am looking for some kind of resolution in this case, otherwise I can not handle this kind of failure which is prone to happen.

Can you please help me. I appreciate your support.

@rbeerak​ 

I happy to let you know that the KafkaConnector does support manual commits now. Once you upgrade to the newest version you will see a new Action called Commit. To use this you add your receiving KafkaConnector setting the Action to Receive and the Auto Commit to No and in the committing KafkaConnector you reference the Receiving Connector Identifier returned by the first KafkaConnector and set your offset and partition.

Let us know if you have any feedback!