Re: [openorb-users] OpenORB NamingService: java.io.EOFException

[<lak...@t-...>]

rosmi.jose wrote:
> Hi,
> I am facing problem with CORBA service. The corba goes down

What do you mean by this? Does the NamingService process stop on the OS 
level, similar to System.exit()? Or do you just see the EOF exceptions 
below?

> frequently in this site and because of that related services also goes 
> down.  Can anyone help me in finding why CORBA is getting restarted in 
> this site? Please provide your inputs on this.
>
> Following are the errors, I could see from the related log files. 
> CORBA.COMM_FAILURE failure error was shown once as shown below.
>
> 07.04.26 05:25:53.303 VRB (MI2Server) [main] 
> GWCAlarmNotifListener::cleanup()-caught Exception, 
> Ex=org.omg.CORBA.COMM_FAILURE: vmcid: SUN minor code: 201 completed: No
>
> Main exceptions thrown are
>
> Error receiving message in process_or_enqueue
> Apr 26 10:32:49 anycmt OpenORB NamingService: java.io.EOFException
> Apr 26 10:32:49 anycmt OpenORB NamingService:   at 
> org.openorb.iiop.IIOPTransport.readMagic(IIOPTransport.java:650)
> Apr 26 10:32:49 anycmt OpenORB NamingService:   at  ...

I think what you see is a timeout on a socket read. In older versions of 
OpenOrb this was logged as an error, but it is actually not. It's 
perfectly normal, the server is waiting for more requests from the same 
client - if no further requests arrive within a certain amount of time 
the server closes the connection to save system resources.

Are you running OpenORB 1.4?

> 07.04.26 05:25:52.593 CRT (TrunkProv@1.0) [main] 
> org.omg.CORBA.OBJECT_NOT_EXIST:   vmcid: OMG  minor code: 2 
>  completed: No
>         at 
> sun.reflect.NativeConstructorAccessorImpl.newInstance0(Native Method)
>         at ...
>         at 
> org.omg.CosNaming._NamingContextExtStub.unbind(_NamingContextExtStub.java:440) 
>

This is also normal. The client tries to unbind an entry from the 
NamingService which does not exist (any more). The server throws an 
OBJECT_NOT_EXIST exception, because such a request can't be handled in a 
meaningful way. Either the calling client code needs to be very sure 
that the object exists before calling unbind, or it needs to handle the 
exception.

If other services go down because of exceptions like this, which system 
causes that behavior? Do you have some monitoring software installed?

Regards,
Lars