Closed Technoboy- closed 5 years ago
confirm this happen in K8s my deployment , but in Controller . When start - Controller list out all instance which connected with. All - i mean the list included old instance which terminated long time ago .
we will discuss if it is the Helix bug. but, take a look at my new pr , that will resolve the controller registry with ip instead of hostname. if you use federate mode, controller will register itself using hostname that will result manager can not request to it via http
this is expected, Instances contains old instance and Live Instances are stored under /LIVEINSTANCES path.
when restart workers in docker, below list has some unlive instances