Multiple CALL dbms.queryJmx("org.neo4j:*") queries slowing down database

[inersha]

My database has gone a bit slow. I checked the currently running queries, and I've got 10 of these running:

CALL dbms.queryJmx("org.neo4j:*")

I didn't call these myself, and they have been running for over 10 minutes each (which is longer than the timeout I have set).

Is there any way to stop these queries? I would kill each of these queries manually, but the database is running so slow I can't even long in to cypher-shell (Unable to establish connection in 5000ms).

Also, what's causing these queries to run?

[andrew_bowman]

These are JMX calls being made from the browser for monitoring purposes.

Can you confirm your Neo4j version?

I believe we had a bug for certain slow JMX queries (being issued by the browser for monitoring) affecting some of our 3.4.x versions, and should have been fixed in 3.4.10.

Please use the latest patch release for the minor version you're running.

[inersha]

Thanks Andrew. I am using neo4j-enterprise 3.5.1. Would upgrading to 3.5.2 fix it?

The browser is Neo4j Browser version: 3.2.15.

Also, is there anything I can do to kill the currently running queries? I've still got 10 of them running, each for 3+ hours.

At first I couldn't log in to cypher-shell because I was unable to establish a connection in under 5000ms. However, I am now able to log-in, but when I try and kill the individual queries I get the following response:

neo4j> CALL dbms.killQuery("query-6285314");
+-----------------------------------------------+
| queryId         | username    | message       |
+-----------------------------------------------+
| "query-6285314" | "me"        | "Query found" |
+-----------------------------------------------+

1 row available after 14 ms, consumed after another 0 ms

Do I have any options other than restarting Neo4j (possibly a hard-kill)?

[inersha · Accepted solution]

This appears to be due to a bit of a serious browser bug. Here is the explanation by Max De Marzi from the Slack channel:

Restart Neo4j with all browser tabs closed.

That jmx query is sent to the server for every query you send via the browser and every 20 seconds.

The jmx query requests everything, it shouldn’t. Fix goes out next Monday.

If the jmx query runs into a heavy write transaction, it asks for which nodes/rels are locked and the lock manager created a billion strings for each one and it basically makes so much garbage memory it kills Neo4j.

Run queries via the shell until the browser update.