Hello,

Our network administrators say that *our company's switches do not block
multicast or broadcast traffic*. To check if this it's true, we've forgotten
about Reggie and we've written a little Java program that simply uses a
MulticastSocket to send packets to the multicast IP *224.0.1.84 and
attacking the port 4160*. Of course, this program failed in the network
office, while it worked perfectly on my home's network.

We are completely sure that it's a problem related to multicast packets.
Maybe it's related to Windows XP or 7 Firewall? In the company there is a
Group Policy commanding the Firewall, but we can define some exceptions and
we have the multicast/broadcast response enabled. Is that enough, or is
there some other issue that remains hidden?

Please, any hint is welcome, because our goal is to deploy this software in
a production environment, not only in a domestic network!

Thanks and regards



2011/9/29 Christopher Dolan <[email protected]>

> Yes, most enterprise switches block multicast by default. That's probably
> your issue.
>
> Another possible issue is reverse DNS. If reggie is broadcasting a private
> hostname or IP address for other machines to call back to it, then it's not
> going to work. We've also had issues with dual-NIC servers where clients
> always try to connect to the primary NIC (as specified in the Windows
> interface binding order) and do not fail over to the secondary NIC. The
> usual cause of that problem is passing a null host to
> TcpServerEndpoint.getInstance() because in that code path, JERI just picks
> the first IP address from a reverse DNS lookup. The solution in that case is
> to instead pass the results of
> InetAddress.getLocalHost().getCanonicalHostName() as the hostname, or
> hard-code the public host name.
>
> I'm suspicious of your "just once" result. Maybe you changed the group
> name? Remember that the group is case-sensitive.
> Chris
>
>
> -----Original Message-----
> From: Sergio Aguilera Cazorla [mailto:[email protected]]
> Sent: Thursday, September 29, 2011 1:16 PM
> To: [email protected]
> Subject: Re: Reggie's visibility in discovery process
>
> Hello,
>
> At last we could perform some testing on Reggie's discovery using multicast
> protocol. I can provide the following results:
>
> - The program performs perfectly when we make a Unicast discovery,
> attacking
> directly the URL where the Lookup service is located.
>
> - We can ping the two machines. Even, we can acces the folder server by the
> HTTP server, and get the reggie-dl.jar classfiles needed to communicate
> with
> the Lookup.
>
> - All machines in the office are Windows XP SP3 and Windows 7. None of the
> combinations server/client XP-7 has thrown a good result.
>
> - We are communicating through switches in the LAN of a enterprise. Do you
> think that multicast packets are bein blocked by intermediate nodes?
>
> - The most misteryous fact: we could perform multicast discovery
> succesfully
> *just once*, the first time we tried. That's suspicious, is there some
> class
> or service that remains hidden and doesn't allow you to perform multicast
> discovery more than once?
>
> Any help on reggie's weird behaviour is very welcome. If I can solve this
> problem, no doubt I will write a short explanation for the community,
> because I think it has to be a very common problem.
>
> Regards.
>
>
>
> 2011/9/22 Иван Бишевац <[email protected]>
>
> > 1. Could you ping two machines?
> > 2. Which operating system you use?
> > 3. Are you communicating through router?
> >
> >
>



-- 
*Sergio Aguilera*

Reply via email to