Kubernetes API connection unstable

Hi, we’ve been investigating Kubernetes running on Rancher but it seems network connections to the apiserver are very unstable. We see sporadic timeouts from services like helm/tiller and kubectl. In the logs we see a number of connection resets e.g.:

2017/11/03 01:30:45 uninstall: Failed deletion of “redacted”: Delete https://172.19.0.1:443/api/v1/persistentvolumes/redacted: read tcp 172.18.176.125:60152->172.19.0.1:443: read: connection reset by peer

E1103 02:34:53.839563 5 leaderelection.go:253] error retrieving resource lock nginx-ingress/ingress-controller-leader-nginx: Get https://172.19.0.1:443/api/v1/namespaces/nginx-ingress/configmaps/ingress-controller-leader-nginx: read tcp 172.18.8.82:39570->172.19.0.1:443: read: connection reset by peer

Does anyone have any ideas on troubleshooting this? We’re using the ipsec network plugin.