Ref #47865 (comment)

From our recent scale test runs:

• With services disabled, size of kube-proxy.log = 921B
• With services enabled, size of kube-proxy.log (+ rotated logs) = ~6 GB

The reason is we are printing the whole iptables (which is super big in our tests) in case of failure in the restore step (this line -https://github.com/kubernetes/kubernetes/blob/master/pkg/proxy/iptables/proxier.go#L1574)
This huge log traffic seems to oom-kill fluentd too.

cc @kubernetes/sig-network-misc @kubernetes/sig-scalability-misc @bowei @gmarek