Already tried restarting the servers. I faced similar issue while i was doing setup with ELB as LB. Hence based on observation it looks like issue with the clustering setup only.
and most important thing ... if there are two instances say 1a and 1b. it works fine if i take either of the nodes offline. This is not an issue with node. It happens when we use clustering mode.
It works for below mentioned scenarios :-
upstream backend {
server sampl2.example.com:443;
}
upstream backend {
server sampl1.example.com:443;
}