Cluster singleton actor lost message stealthily

my cluster have 2 node in one machine,when i tell some message to actor ‘a’ (singleton actor at Node2) continually at Node1,in my test,i found send 5k message will lost 800+!!!!all config is default, and the actor a just count the message amount and do nothing.

Akka actor messages are inherently unreliable, this is intentional design.

It also depends on how you test this. For example, if you send 5k messages in one go you will probably overflow some buffers and messages will be dropped. If you don’t have any other failures, such as restarting Node 2 and you send them in “normal” pace there will not be any lost messages.