CODEWITHSUNDEEP
javarmi
ParagJ
ParagJ

Reputation: 1582

JDK7u21 RMI changes

The Release Notes of JDK7u21 specified the changes related to RMI:

From this release, the RMI property java.rmi.server.useCodebaseOnly is set to true by default. In previous releases the default value was false.

This change of default value may cause RMI-based applications to break unexpectedly. The typical symptom is a stack trace that contains a java.rmi.UnmarshalException containing a nested java.lang.ClassNotFoundException.

If these exceptions occur, the preferred way to solve the problem is to configure all RMI clients and servers to use the same codebase, by specifying proper values in the java.rmi.server.codebase system property. This is typically done by adding the -D option to the command that starts up the application: java -Djava.rmi.server.codebase=file:////(path-to-remote-classes)/

It also specified in the documentation here that workaround is to set the java.rmi.server.useCodebaseOnly property back to false.

However, even when I explicitly set this property to false, I am getting an exception mentioned above. Any clue here please?

And if I have to solve it by the preferred way described above (by setting the -Djava.rmi.server.codebase=file:////(path-to-remote-classes)/, then which path should I set over here? My workspace/bin?

Upvotes: 1

Views: 1234

Answers (3)

ParagJ
ParagJ

Reputation: 1582

I am sorry, I solved it on my own. The problem was wrong number of parameters were being passed to my .ksh file. This caused the problem starting RMI registry.

Upvotes: 0

user207421
user207421

Reputation: 310869

You would need to either:

  1. Set it to false at all the JVM(s) which download classes, and ensure that java.rmi.server.codebase is set to something useable at all the JVMs which send objects whose classes may need downloading, or

  2. Leave it alone but make sure that every JVM in the system is started with a java.rmi.server.codebase property that points to a useable codebase for that JVM.

It's a very strange change to make. I don't see any security implications, just a nuisance change that changes semantics at all the codebase recipients. With useCodebaseOnly=true, you have to set java.rmi.server.codebase at the receiving JVMs, where otherwise you only have to set it at the sending JVMs. This isn't clear from the documentation.

Upvotes: 0

Stuart Marks
Stuart Marks

Reputation: 132350

You have to make sure to set the properties on all JVMs that are using your remote classes. This includes your RMI server, the RMI clients, and the RMI registry. If they are all on the same machine, then you can set java.rmi.server.codebase on them all to a file: URL that points to the location of the classes. If they're not on the same machine, then you could make the classes available using an HTTP server, and then set the codebase property of the remote JVMs (presumably the clients) to use an http: URL that points to where the classes are available. Or you could make the classes available to remote JVMs through some other means, such as copying them, or using a shared filesystem, and then set the codebase property to a file: URL.

Your attempt to apply the workaround of setting java.rmi.server.useCodebaseOnly to false might not be working for a couple reasons: you might not have set it in all of the interacting JVMs, or you might not have set the codebase property to the right value.

Since you're having to go to the trouble of configuring the registry and all the clients, you might as well pursue setting their codebase property to the right URL, instead of pursuing the workaround of setting useCodebaseOnly to false.

Upvotes: 2

Related Questions