Nikolay Seliverstov

Hi, it seems with 2.12.9 it works quite well. There are no spikes in memory consumption, but still there are slow leaks. Currently it's around steady 13 mb per day....

Unfortunately it's still leaking unexpectedly. Last restart of DB was on 17th Oct (11 days between restart and the leak) 

We have noticed on our side that leaking starts when chef agent runs his check once an hour (the orange spikes). So, maybe that's a factor we need to take...

Well, original query is written to mimic all-in-one UPSERT query. As I understand your query 1. It will not create nodes, I have to create them separately. 2. `MERGE (af)-[f:Edge...

Thanks for the detailed explanation. I have SQL background so that behavior is unexpected for me. Probably I missed that part in the Cypher docs. > If you like I'll...

I have found that leaking happens after internal RedisGraph cleanup cycle.  Now we have much larger graph and it leaks quite often, several times a day spreaded over large...

Good catch. That was my assumption. I forgot we have DELETE query to clean up old data once an hour. So, the leak happens on (or close to) `MATCH ()-[f:Edge]->()...

By the way we still use unoptimized "upsert" query if it's related.

We applied the config change a day ago an do not see changes in memory leak frequency 

That's quite interesting. Today it's chill and no leaks. Many thanks, @swilly22!