These are chat archives for akkadotnet/

Oct 2017
Arsene T. Gandote
Oct 06 2017 07:43
Hello can someone educate on NBench and probably what is its use and a tutorial.
Natan Vivo
Oct 06 2017 15:22
What is the best way to detect if the node is unreachable from within the unreachable node? Will it see the other nodes as unreachable or does it set any status internally to detect this?
I was actually wondering is there is a way to completely avoid the idea of unreachable nodes, and make it just reconnect. I know currently the cluster algorithm depends on that but I'm coming to the conclusion it's simpler to not have these controls and deal with the other issues than deal with unreachable nodes
The fact is that the cloud is a very unstable place, and as the number of nodes go up, I'm having unreachable nodes twice a day now due to conditions I cannot control. These nodes eventually come back, but while that is happening the cluster gets unusable.
Natan Vivo
Oct 06 2017 15:50
Created a discussion on #3144
would like to hear some thoughts on that
Bartosz Sypytkowski
Oct 06 2017 19:45
@nvivo if you think of cluster as a graph, unreachability is a property of a graph's edge, not a node itself (nor a subset of the graph)
in general if A sees B as unreachable, B should see A as unreachable too, but this may not happen at exactly the same moment
also AFAIK unreachable nodes are trying to reconnect to each other every 5s (by default)
also #3099 should help a little - it introduces an extra node status, that will allow you to use nodes even before cluster reaches full convergence (which is hard to achieve when node churn is big)