provide an rpc-proxy

[hyperthunk]

Often you need to rpc to a remote node. Sometimes, operations on that node will break a connection (e.g., RabbitMQ Clustering force-reset commands, which effectively net_kernel:disconnect(Node) for all of nodes()) an in this case the systest process will see a down and puke - this isn't what we want.

Instead, we should either

1. not use net_kernel:monitor_nodes in systest_proc or
2. provide an rpc-proxy so we're never in nodes to begin with

Or both....

[hyperthunk]

Actually it's a bit simpler to do (2) and I did prototype this, though it'll need writing properly again.

[rade]

Why doesn't systest connect as a hidden node? Well, according to hyperthunk it actually does, but somehow the rpc calls systest is making ends up connecting non-hidden. That is strange since rpc isn't doing anything unusual.

[hyperthunk]

Indeed I'm not convinced by the validity of this issue any more.