Problem with left-join

Kafka Streams
1

Hi guys I have a problem with KStreams. I have a KTable T1 on which I perform a left join on a KTable T2.
The result is converted to a KStream.

I would like for each update in table T2 the left-join to publish a message. But this does not always happen.

Sometimes I am forced to do a dummy update of T1, then later I can update T2 as much as I want and get the behavior I want.

Is it basically wrong to expect a join update when only the aggregate table is changed?

I also attach a piece of code to make the problem clearer.
ProductFacility is the table I would like to update.

    kStream = productKTable.getKTable()
            .joiner()
            .leftJoin(
                    productFacilityKTable.getKTable(),
                    (product, productFacilities) -> {
                        if(productFacilities != null){
                            var facilities = new HashMap<String, ErpProductFacility>();
                            for (var productFacility : productFacilities) {
                                facilities.put(productFacility.facility, productFacility);
                            }
                            product.facilities = facilities;
                        }
                        return product;
                    },
                    ElMat.ephemeral()
            )
            .toStream()

Any help will be appreciated. Thanks in advance.

2

KTable by default use internal caches. You might want to disable caching, either globally for all KTables setting cache size to zero in the config, or on a per-table basis (for both input tables) by passing in a Materialized config object.

Cf. Kafka Streams Memory Management | Confluent Documentation