Okay, so here is what I have now. I am setting up an icecast server for our
customers
to use for streaming audio -- in particular there will be a couple of radio
stations
in town using it. I want it to run on a cluster of systems for redundancy. So I
have
the server running on two identical machines on one ip that is dedicated to
streaming
services, behind a foundry switch. The foundry switch directs connections to the
standard ports according to which has the least connections.
Each machine also has another port that is bound to the ip of the machine itself
and
is just for me to specifically connect to each machine, and also for them to
talk to
each other. Each machine is set to use the other as a master. So when someone
connects to send a stream, whichever machine they connect to, the other machine
picks
up the mountpoint when it scans the master.
This works, except for one thing. I have to test with it some more, but it seems
like
if something causes the sending client to disconnect, so that the initial
machine
loses the mountpoint, it will then pick the mountpoint up from the other machine
when
it scans it. Then when the sending client reconnects, it gets refused saying the
mountpoint is already in use. The two cluster machines then hand the mountpoint
back
and forth to each other.
I need to test with it some more, there may be something I can do with update
times
to do something about this.
But is there a better way to accomplish the same goal? I am trying to create a
redundant streaming server, basically.
Joe