IPv6 wrongly enabled, causes "Cannot assign requested address"

[antoineco]

NGINX Ingress controller version: master @ a688208

Environment: Docker for Mac

• OS: Linux
• Kernel: 4.9.93-linuxkit-aufs

What happened:

net.IsIPV6Enabled() wrongly returns true because the /proc/net/if_inet6 file exists in the container, however my network interfaces do not have IPv6 addresses:

$ kubectl -n ingress-nginx exec nginx-ingress-controller-8547fdb4d9-f94n4 -- ip a
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1
    link/loopback 00:00:00:00:00:00 brd 00:00:00:00:00:00
    inet 127.0.0.1/8 scope host lo
       valid_lft forever preferred_lft forever
2: ...

As a result, (re)loading the NGINX config fails with

nginx: the configuration file /tmp/nginx-cfg255014169 syntax is ok
2018/08/06 10:00:51 [emerg] 50#50: bind() to [::1]:18080 failed (99: Cannot assign requested address)
nginx: [emerg] bind() to [::1]:18080 failed (99: Cannot assign requested address)

What you expected to happen:

$IsIPV6Enabled is false when network interfaces can not handle IPv6, and NGINX does not try to listen on IPv6 addresses.

How to reproduce it:

Simply run the Ingress controller in an environment where the container's network interface do not have IPv6 addresses. In the case of Docker for Mac, --network-plugin=cni and an empty --cni-conf-dirdirectory (kubelet flags).

/kind bug