Your update to the KTable is async when you send data back to the KTable
input topic. So your program is subject to race-conditions.
So switching to the PAPI was the right move: it make the update to the
state store sync and thus fixes the issue.
-Matthias
On 9/4/23 5:53 AM, Mauricio Lopez
Hello,
They were getting processed by the same consumer as we only had a single
machine running this.
What we ended up doing is basically drawing the same topology but interacting
directly with the stateStore using the Processor API instead of DSL. Seems that
fixed everything up (and made it
Hi,
I'm definitly no expert, but to me it sounds as not all your messages are
getting processed by the same consumer. Are you using the key `foo` for
partitioning? Is `baz` actually another key or is this mixup in your example
and `baz` is another value with key `foo`?
Hope you find a
