On Wednesday, September 11, 2013 12:42:33 PM Paul Colomiets wrote: > Hi Martin, > > Just a couple of comments. > <snip> > > 3. The priorities stuff is complex to setup. Note there are only three > > > >> low priority connections on the scheme that make whole cluster to be > >> switched when no workers available. Note also that the only priorities > >> do prevent messages being bounced in the cluster forever (in the case > >> there are at least single worker). All in all there are many tiny things > >> that is complex to setup unless there are appropriate tools. > > > > So what exactly do you feel is complex about prioritised connections? Loop > > avoidance? > > Probably it's gut feeling. I would say I've failed to do low priority > connections in that topology, unless I see them on the graph. But drawing a > graph for 1000 nodes (each have tens of processes), is not going to work > either. Well, what about annotating some points on the graph to be _roles_ rather than individual nodes? With REQ/REP for instance, due to its stateless nature, the 'endpoints' could describe a _class_ of nodes rather than individuals? So for instance, you could have A(REQ) -> B(device) -> C(REP), but A and C are annotated as roles. Then _any_ node can join as 'A' or 'C', get the appropriate configuration for that _role_, and life goes on.