[CASSANDRA-18583] Fix shutting down IsolatedJmx

• Type: Bug
• Status: Resolved
• Priority: Normal
• Resolution: Abandoned
• Fix Version/s: 3.11.x, 4.0.x, 4.1.x, 5.x
• Component/s: Test/dtest/java
• Bug Category: Code
• Severity: Low
• Complexity: Low Hanging Fruit
• Discovered By: Adhoc Test
• Platform: All
• Assignee: Stefan Miklosovic
• Reporter: Stefan Miklosovic

Authors

Stefan Miklosovic

• Created: 11/Jun/23 20:43
• Updated: 12/Jun/23 15:39
• Resolved: 12/Jun/23 15:39

Description

public class MyTest extends TestBaseImpl
{
    @Test
    public void test() throws Throwable
    {
        try (Cluster cluster = Cluster.build(2).withConfig(c -> c.with(Feature.values())).start())
        {
            ClusterUtils.stopUnchecked(cluster.get(2));
            System.out.println(cluster.get(1).nodetoolResult("status").getStderr());
        }
    }
}  

This will fail on this:

java.net.BindException: Address already in use (Bind failed)
	at java.net.PlainSocketImpl.socketBind(Native Method)
	at java.net.AbstractPlainSocketImpl.bind(AbstractPlainSocketImpl.java:387)
	at java.net.ServerSocket.bind(ServerSocket.java:390)
	at java.net.ServerSocket.<init>(ServerSocket.java:252)
	at javax.net.DefaultServerSocketFactory.createServerSocket(ServerSocketFactory.java:231)
	at org.apache.cassandra.distributed.impl.CollectingRMIServerSocketFac

The problem is that IsolatedJMX clears whole map of TCPTransports in clearMapField as part of the stopJmx method. This results in JMX internals thinking the socket is not there so it tries to create it but it fails to do so because it is technically still bound.

The fix is consisting of cleaning just the bits belonging to that specific IsolatedJMX instance.
The investigation of the issue done by myself, the actual fix by drohrer. Found while working on CASSANDRA-18572.

Issue Links

• blocks: CASSANDRA-18572 Instance.nodetoolResult should connect to JMX if there is such feature enabled in its config