[Spread-users] single vs. multiple daemons
yairamir at cnds.jhu.edu
Thu Sep 20 16:35:52 EDT 2001
I think the place to start is to run the spuser - an application provided with
Spread, and see what you get. This will eliminate application
issues and show the real Spread picture. You can then join a group
with multiple users and see what happens. You can check it with one
daemon or with many daemons. If all of this works, it is likely you
have a problem with your application. If it does not work, it is likely
you have a problem with your setup. If both are ok, then maybe there
is a bug but I can hardly believe it - because such a bug will be visible
In that case, e-mail your OS version, spread.conf, and the specific
scenario with spuser that shows the system is broken.
If your OS is one of these we provide executables for, I would appreciate
you also running the provided executables rather than the ones you build.
I am sure this can be traced quite quickly.
Roy Tromble wrote:
> I'm testing an application using Spread and encountering some bothersome
> behavior. I have three nodes running on virtually identical machines in
> different subnets. Another node sends requests to these three via Spread
> and gets responses back from them to its private group. Here is the
> problem: if I run a separate Spread daemon on each machine, and have the
> nodes connect to that daemon, their responses are different from when all
> three connect to a single Spread daemon. If they all connect to one, I get
> the behavior I'm expecting. However, if they connect to multiple daemons, I
> get responses that have bad data in them. I can't characterize the bad
> data, because I'm not sure what's wrong with it (it isn't printable), but
> the application handles it incorrectly. Also, it is not consistently bad.
> Often during the different cycles, two of the three will respond correctly,
> while during other cycles the other will respond correctly.
> I'm wondering if there is any reason why the three would have communication
> problems. I'm pretty confident that they have the same configuration and
> are using the same version of Spread.
> Roy Tromble
> Spread-users mailing list
> Spread-users at lists.spread.org
More information about the Spread-users