<https://stackoverflow.com/questions/79758622/how-...
# troubleshooting
m
v
Do what you do in 1st step. and use setting to not start from savepoint. that it will start from earliest offset
p
Do you also want to keep the curent state or do you want to rebuild state from scratch? If you want to create new state, the first option described could work. You just need to start fresh and ignore the current state. If previous state is important to you, I'd create a new Kafka source with different naming/uid to drink the data. Therefore Flink will understand that as a new piece. The issue is that Flink does not rely on kafka consumer groups. It relies on its own mechanism of offsets.
m
I need to start from savepoint for some insert statements. I just need to avoid starting from savepoint for this insert statement alone,is there a way to do that in flink sql?
1. I dont want to keep the current state,i need to rebuild state from scratch. 2. The first option did not work. 3. My kafka source is a table in flink sql, i think uid is not possible with flink sql. Is there any other way i could try to reset the offset for just a single source?
e
If it's a one time thing, you could theoretically just write a python script that uses the same credentials as your Kafka source, and just change the offset. Just make sure that your flink app is not running at the same time so the Kafka brokers will accept the offset resetting. If the savepoint also saves the last offset and flink uses this offset no matter what, you can try and just change the kafka credentials at each startup of your app Can you explain why you need mechanisms for resetting the offset while using the same savepoints and state?
m
If i add an extra column in,or if i want to change the population logic of a column in flink sql joins i want to reset the offset.