Pregel function call never ends - Although log says it's finished

Neo4j Graph Platform
1

I wrote a simple pregel code to find shortest path between given source,target pair. However when I call the function via cypher the execution never ends. I looked at the log and it lists iterations and after that it says it's finished. When I force terminate the query an exception is thrown: Here is the exception :

ERROR [o.n.b.r.DefaultBoltConnection] Protocol breach detected in bolt session 'bolt-0'.
org.neo4j.bolt.runtime.BoltProtocolBreachFatality: Message 'PULL Map{qid -> Long(0), n -> Long(1000)}' cannot be handled by a session in the READY state.
at org.neo4j.bolt.runtime.statemachine.impl.AbstractBoltStateMachine.nextState(AbstractBoltStateMachine.java:159) ~[neo4j-bolt-4.4.7.jar:4.4.7]
at org.neo4j.bolt.runtime.statemachine.impl.AbstractBoltStateMachine.process(AbstractBoltStateMachine.java:102) ~[neo4j-bolt-4.4.7.jar:4.4.7]
at org.neo4j.bolt.messaging.BoltRequestMessageReader.lambda$doRead$1(BoltRequestMessageReader.java:93) ~[neo4j-bolt-4.4.7.jar:4.4.7]
at org.neo4j.bolt.runtime.DefaultBoltConnection.lambda$enqueue$0(DefaultBoltConnection.java:156) ~[neo4j-bolt-4.4.7.jar:4.4.7]
at org.neo4j.bolt.runtime.DefaultBoltConnection.processNextBatchInternal(DefaultBoltConnection.java:252) ~[neo4j-bolt-4.4.7.jar:4.4.7]
at org.neo4j.bolt.runtime.DefaultBoltConnection.processNextBatch(DefaultBoltConnection.java:187) ~[neo4j-bolt-4.4.7.jar:4.4.7]
at org.neo4j.bolt.runtime.DefaultBoltConnection.processNextBatch(DefaultBoltConnection.java:177) ~[neo4j-bolt-4.4.7.jar:4.4.7]
at org.neo4j.bolt.runtime.scheduling.ExecutorBoltScheduler.executeBatch(ExecutorBoltScheduler.java:257) ~[neo4j-bolt-4.4.7.jar:4.4.7]
at org.neo4j.bolt.runtime.scheduling.ExecutorBoltScheduler.lambda$scheduleBatchOrHandleError$3(ExecutorBoltScheduler.java:240) ~[neo4j-bolt-4.4.7.jar:4.4.7]
at java.util.concurrent.CompletableFuture$AsyncSupply.run(CompletableFuture.java:1700) [?:?]
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1128) [?:?]
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:628) [?:?]
at io.netty.util.concurrent.FastThreadLocalRunnable.run(FastThreadLocalRunnable.java:30) [netty-common-4.1.75.Final.jar:4.1.75.Final]
at java.lang.Thread.run(Thread.java:829) [?:?]

I have a distance variable in my schema(which is public) and I YIELD the call to function with 'YIELD values'.

What's the problem?

Thanks.

2

Apart from what martin wrote, it would also be helpful if you could share your pregel example or at least some code for us to reproduce the error.

Cheers

3

Could you please open an issue on https://github.com/neo4j/graph-data-science

4

So as @martin_junghann pointed out in https://community.neo4j.com/t5/neo4j-graph-platform/pregel-node-degree-is-0-even-if-it-has-relationships/td-p/15815/page/2 the node ids obtained from config and context don't match because internal gds ids are different than neo4j database ids. I campare these two. That's why computation never ends - it never converges. But I took that code from pregel examples. There must be something wrong with those examples as well. And in this case, how should we find the startnodes or end nodes in our algorithm during procedure call?

Also, still, even though max iterations has been reached and log says 'Finished', neo4j browser still stucks at loading screen. No exceptions.

5

I'm opening the issue.

6

Here is my code: https://gitlab.com/cuneyttyler/dijkstra-multiple-inputs

7

As discussed on the GitHub issue we don't consider this as a bug, since the user attempts to stream 10 million result rows which just takes time. The "Finished" log refers to the computation of the algorithm and not to the point in time where the results are visible in e.g. the browser.

To address the Pregel node id space issue, as discussed on the other thread, I added functionality to read the original node id from the Pregel compute context. I updated the documentation and also the example code . It's released as part of GDS 2.1.