David Allen

Underlying implementation is here: https://github.com/neo4j-contrib/neo4j-streams/blob/8f986d684730eafdde3617768e99ddb02ddc72f4/consumer/src/main/kotlin/streams/procedures/StreamsSinkProcedures.kt#L50 Kernel team has indicated that consumption or non-consumption of the results should not affect whether the TX executes or not.

@ivangreene starting the sink indeed does have different threading consequences, which is a conflating factor here, but you can't really move that call outside of the path. Otherwise if starting...

This will likely require us to repro and do some investigation. Good enough for the moment that the work-around is so trivial (just consume the results)

@conker84 closeable by previously merged tickets?

@mroiter-larus please try again with bolt+routing://. The default connection URI for Aura is indeed neo4j:// and this does require a java driver update in order to make possible (if we...

@conker84 OK, in this case, this validation does need to be expanded. The current neo4j drivers manual includes a number of different variants, including bolt+s, bolt+ssc, etc. Be careful --...

@Tin-Nguyen thanks for the feedback - seems this is a priority for you too, please add any detail that you can on what kinds of crashes you're seeing to help...

I recently talked to @rhaces on the phone, and suggested some more architectural fixes for this rather than better enabling people to write really large cypher queries. Please leave this...

This is an important ticket, but we need more internal team discussion on how to move forward with it. The issue is that what users really need is a work-around...

Related https://github.com/neo4j-contrib/neo4j-streams/issues/181