[Spread-users] Using spread in secure group communication

[John Schultz]

Hi Adam,

The first thing to do is to let us see the actual configuration file that 
you are using.  Usually problems like this are related to simple 
configuration errors.

Your understanding of Spread's communication is *mostly* correct. 
However, within a segment it is understood that when a daemon does a 
hardware multicast that all the other daemons in the segment will have a 
non-zero (actually quite high) chance of hearing the multicast.  That is 
in fact the exact definition of a network segment in Spread.  So within a 
segment it is generally assumed that you have full connectivity between 
all the daemons there.  When this isn't the case, Spread will sometimes 
have trouble functioning or be forced to do extra work in pathological 
manner (e.g. - always recovering msg "losses").

The fact that none of your daemons are perceiving each other hints to me 
that you either have a configuration problem or a bad network setup (e.g. 
- no multicast or broadcast).

As to your question on dynamic reconfiguration of Spread daemons, the new 
version of Spread, which will be going into public beta *very* soon has 
added exactly that feature.

Cheers!

---
John Schultz
Spread Concepts
Phn: 443 838 2200

On Mon, 13 Feb 2006, Adam Lamar wrote:

> Hello,
>
> We are considering using spread for a secure group communication application. 
> Spread seems to be the perfect fit, however, we are having trouble getting 
> spread to work exactly as we understand it does.
>
> We are testing spread on four machines on our local network. Each machine has 
> the spread daemon running and can communicate directly with every other 
> machine. We have defined our spread segment in the spread.conf file with a 
> listing of each machine name and IP address. Communication is working with 
> the spuser program in the client/server architecture, i.e. if machines 
> A,B,C,D connect to machine A's spread daemon, each of the machines is able to 
> join the group and send/receive messages.
>
> The problem arises when we try to use spread in a distributed nature. If each 
> machine is running the spread daemon, it is our understanding that machine A 
> can connect to B, B can connect to C, and A will be able to communicate with 
> C through B. However, when we tried doing this with the spuser program, this 
> communication does not occur. In fact, watching the packet dump does not 
> reveal that any communication whatsoever is occuring between the spread 
> daemons. Is there something obvious that we are doing wrong?
>
> Also, as we plan on using spread for dynamic groups, is is possible to define 
> (or for spread to automatically find) where each spread daemon is located 
> without restarting the spread daemon each time a new participant joins the 
> network?
>
> Thank you,
> Adam Lamar
>
> _______________________________________________
> Spread-users mailing list
> Spread-users at lists.spread.org
> http://lists.spread.org/mailman/listinfo/spread-users
>