Query execution threads are not interrupted/terminated when the driver is closed (neo4j driver version 5.9.0 and above)

[aevangelista81]

Hi guys,
It seems that closing a driver via the Driver.close() method from another thread than the one owning the driver instance will not interrupt long running transactions busy on the owning thread. The issue also shows that the owning thread will never finish, even after the transaction query should have been completed.

Probably a piece of code is better to explain the behaviour :

package org.example;

import org.neo4j.driver.AuthToken;
import org.neo4j.driver.AuthTokens;
import org.neo4j.driver.Config;
import org.neo4j.driver.Driver;
import org.neo4j.driver.GraphDatabase;
import org.neo4j.driver.Result;
import org.neo4j.driver.Session;
import org.neo4j.driver.SessionConfig;
import org.neo4j.driver.TransactionConfig;

import java.time.Duration;
import java.util.Timer;
import java.util.TimerTask;
import java.util.concurrent.TimeUnit;


public class Main {

    static Driver driver;

    public static void main(String[] args) {
        System.out.println("PID: " + ProcessHandle.current().pid());


        Config config = Config.builder()
                .withMaxConnectionLifetime(30, TimeUnit.MINUTES)
                .withMaxConnectionPoolSize(50)
                .withConnectionAcquisitionTimeout(2, TimeUnit.MINUTES)
                .build();

        String boltUri = "bolt://localhost:7687";
        AuthToken authTokens = AuthTokens.basic("neo4j", "password");
        driver = GraphDatabase.driver(boltUri, authTokens, config);
        final Timer timer = new Timer("Closing Driver");

        final TimerTask task = new TimerTask() {

            @Override
            public void run() {
                System.out.println("Closing Driver");
                //Here I'm excepting that the thread running the query (main application thread) will be interrupted and this program will finish the execution, but this will never happen.
                driver.close();
                timer.cancel(); // stop timer after execution

            }
        };
        timer.schedule(task, 10000);

        SessionConfig sessionConfig = SessionConfig.defaultConfig();

        try (Session session = driver.session(sessionConfig)) {
            session.executeWrite((tx) -> {
                //Simulate a long running transaction
                Result result = tx.run("call apoc.util.sleep(30000)\n" + "MATCH (n) RETURN n");
                result.forEachRemaining(r -> {
                    //Do nothing
                });
                return null;
            }, TransactionConfig.builder().withTimeout(Duration.ofMinutes(2)).build());
        } catch (Exception e) {
            System.out.println("Error: " + e.getMessage());
            throw new RuntimeException(e);
        }

        System.out.println("Finished main thread");
    }
}

In this specific case the previous application will never end.

Thanks in advance,
Andrea Evangelista

[ikwattro]

👀

[ikwattro]

For info, this issue appears since version 4.2.2, downgrading to 4.2.1 shows the test above do what's expected.

[AndyHeap-NeoTech]

Thank you for the bug report. We will start an investigation into this.

@ikwattro Is it an option to upgrade the driver? 4.4 is our LTS version, and the latest is 5.9. We recommend upgrading to the latest version whenever possible. Any fixes required will go onto the 4.4 LTS and the next release of the 5.x driver.

[ikwattro]

As mentioned in the title, it happens on 5.9, we did reverse on all versions until a version that doesn’t have this bug and we can confirm the 4.2.2 version of the driver introduced this bug. We are on 5.9 with a custom patch for now, so it’s not urgent although we prefer not to use custom builds of the driver for too long. Cheers !

On Wed, 28 Jun 2023 at 17:46, Andy Heap ***@***.***> wrote: Thank you for the bug report. We will start an investigation into this. Is there an option to upgrade the driver? 4.4 is our LTS version, and the latest is 5.9. We recommend upgrading to the latest version whenever possible. — Reply to this email directly, view it on GitHub <#1436 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAJKK6OUAFSZ3NBNUKBLBP3XNRGW3ANCNFSM6AAAAAAZUJKZAU> . You are receiving this because you commented.Message ID: ***@***.***>

-- Sent from my iPhone.

[AndyHeap-NeoTech]

Fix now in place and released in driver version 5.10

[ikwattro]

Thanks a lot for the fast resolution !

[injectives]

Thanks!

The issue should be fixed in 5.10.0 and 4.4.12.

[aevangelista81]

Thanks a lot, I can confirm the 5.10.0 release fixes the issue.