[omniORB] CORBA call issues

[Duncan Grisby]

On Wednesday 15 August, YaKun Wang wrote:

> I am working on a large scale closed system (not connected to the internet)
> that does not use a DNS; instead it uses preconfigured host files to
> minimise the impact of network failures. It is possible that these host
> files may become out of sync (although this needs to be managed on the
> production system).
> 
> I have met a problem where CORBA calls fail transiently. The calls originate
> from an IOR-format reference.
> 
> I need to find the answers to the following questions so I can implement a
> solution:
> 
> 1.  Is the location information encoded in an IOR a hostname or an IP
> address? 

By default, omniORB uses the IP address. You can change that with
various endPoint options as described in the manual.

> 2.  When initiating a CORBA call, the _narrow function is used. Are there
> any modifications to location information inside IOR by this _narrow
> function?

No.

> 3.  How does CORBA find the location through this IOR-format reference,
> looking up DNS to find the IP address or other methods? 

If it contains an IP address, which it usually does, there is no
lookup. If it contains a name, it uses the platform's getaddrinfo() or
gethostbyname() function.

> 4. How can I reduce the occurrence of this CORBA call error?

You haven't said how the calls fail. What exactly is the error you get?
Have you run with omniORB's tracing on?  Try running with traceLevel 25
traceInvocations 1 to see what's happening.

Cheers,

Duncan.

-- 
 -- Duncan Grisby         --
  -- duncan at grisby.org     --
   -- http://www.grisby.org --