After start-up we see an out of memory condition occurring within 24hrs usually, depending on max heap size set. We have been taking period heap dumps using map. Using the Eclipse Memory Analyzer we can see the following info on a leak suspect: One instance of ‘org.neo4j.bolt.transaction.StatementProcessorTxManager’ loaded by ‘jdk.internal.loader.ClassLoaders$AppClassLoader @ 0xe08f5ea0’ occupies 311,070,080 (90.32%) bytes. The memory is accumulated in one instance of ‘java.util.concurrent.ConcurrentHashMap$Node[]’, loaded by ‘’, which occupies 311,069,848 (90.32%) bytes.
Versions:
Neo4j community (single server) version: 4.4.11 and 4.4.10 Operating system: Ubuntu 22.04.1 LTS running on Azure App Service for Linux as Docker container API/Driver: Neo4jClient 4.1.26 (HTTP client) .NET Version .NET core 3.1
We have tried larger heap sizes(currently max of 512Mb) but we just see the time before failure increasing and the memory used by org.neo4j.bolt.transaction.StatementProcessorTxManager increasing. All queries (no writes) are sent using a single request against the /db/neo4j/tx/commit (by the .Net C# client).
Looking for ideas as to whether this is a database, client or configuration / usage issue that others might have seen and know how to resolve.