MWE displaying a concurrency bug in the Neo4j/gremlin/tinkerpop stack. After a vertex has been deleted, other processors may attempt to access that vertex, resulting in a org.neo4j.graphdb.NotFoundException caused by a org.neo4j.kernel.api.exceptions.EntityNotFoundException.
I suspect that this MWE will run without error on a single-processor machine, but I haven't actually tried that.
To run from the project root directory,
neo4j-deletion-error$ sbt run
To compile and run from elsewhere, first do
neo4j-deletion-error$ sbt assembly
This will create a runnable JAR in target/scala-2.11/ named something like neo4j-deletion-error-assembly-0.1.0.jar. This jar can now be renamed or moved wherever you'd like it.
The only command-line argument that this program accepts is
-s [storageDirectory]
This arg is optional; the default location is HOME/.neo4jbug/db.
This program is configured to log at a fairly chatty level to both the console and to a logfile located at HOME/.neo4jbug/neo4jbug.log. This can all be changed by modifying src/main/resources/logback.xml. A logfile generated by the current master can be found at this Gist.
The observed bug could be anywhere in the Neo4j stack
- gremlin-java
- Apache Tinkerpop API
- Neo4j itself
We need to peel layers off the top of the stack until we find the actual source of the bug.