I'm running the leader-elector (v0.5) as a sidecar on three pods, on three different nodes.

Arguments: --election=XXX --http=0.0.0.0:4040

All works well until I kill the leader pod.

Now I get one of the pods into a state where the logs say it switched to the new leader:

kubectl logs -c elector upper-0
I0803 21:08:38.645849       7 leaderelection.go:296] lock is held by upper-1 and has not yet expired

So that indicates that upper-1 is now the leader.

But if I do query the HTTP server of upper-0, it returns the old leader:

kubectl exec -it upper-0 bash                                                        
root@nso-lsa-upper-0:/# curl http://localhost:4040                                                                                     
{"name":"upper-2"}

Do I need to do something for the leader-electors HTTP service to update?