The ECK operator exits because it fails to connect to the Kubernetes API server. I assume you are proxying these connections with Istio? If so, I would start looking on the Istio side why the connections are being blocked, are the proxies healthy? or are these genuine failures of the Kubernetes control plane then I would start looking into the health of the Kubernetes API server. You posted a similar question about failed leader elections here ECK Operator - Leader Election Lost causes Restart - #2 by pebrc that points to the same underlying problem.