Tom Gordon - 2004-08-10

Hi.  The documentation says to take care not to create cycles when setting up a cluster.  However, it also says that duplicates will be discarded.  If dupes are discarded, what is the harm in creating cycles?  Is is just a matter of additional bandwidth & cpu?

My real question is:  how can I set up a cluster such that if any one node goes down, all other nodes will still be conntected?  Is that possible?  In the documentation example:

A <-> B <-> C

If B goes down then I am hosed.  This is still a case of single failure instead of increased reliability it seems to me.  I really need redundancy in a large network.