Closed arjan closed 5 years ago
I'm trying to think if their may be a better way to do this.
Yeah.. well I have thought about this as well and I think that maybe the whole point of blacklisting / whitelisting nodes should be that you should not use any Swarm function on those nodes.
So maybe this actually does not need a solution. We should focus instead on #104 :-)
I already wrote the proxy solution a while back by the way. But it's hard to get the tests passing with this code, for some reason.
Yeah.. well I have thought about this as well and I think that maybe the whole point of blacklisting / whitelisting nodes should be that you should not use any Swarm function on those nodes.
Yeah that was kinda the way I was leaning too
We should focus instead on #104 :-)
That is going to require some significant work. I'd be more interested in us banging out the conversion to ex_unit_clustered_case for testing so as we move into #104 we have a working test suite to hopefully catch things.
Yes, agreed!
See amongst others #92 and #31. Calling register_name, whereis_name etc. on a node that is blacklisted (or not whitelisted) gives inconsistent and wrong results.
A proposed solution is to not start a tracker process on these nodes, but instead, start a process which proxies all calls (using
:rpc.call
) to one of the tracker processes on a participating node.@beardedeagle how does this sound?