network-get seems different between aws/gce; gives fan interface on gce

[Kevin W Monroe]

With latest/stable juju 2.8.7 (14932), I'm unable to deploy charmed-kubernetes to google, but it's fine on amazon. I think this is because 'network-get' sees multiple interfaces (fan + non-fan) on aws, but only-fan on gce.

Reproduce with 'juju deploy charmed-kubernetes' to both clouds.

On AWS, note multiple bind-addresses for a flannel unit:
-----
$ juju run --unit flannel/0 'network-get cni'
bind-addresses:
- macaddress: 12:6e:dc:6c:6f:15
  interfacename: ens5
  addresses:
  - hostname: ""
    address: 172.31.79.245
    cidr: 172.31.64.0/20
- macaddress: de:1d:bf:6a:7a:8b
  interfacename: fan-252
  addresses:
  - hostname: ""
    address: 252.79.245.1
    cidr: 252.64.0.0/12
egress-subnets:
- 172.31.79.245/32
ingress-addresses:
- 172.31.79.245
- 252.79.245.1
-----

With the following interfaces being available on flannel/0:
-----
$ ip a
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
    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
    inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
2: ens5: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9001 qdisc mq state UP group default qlen 1000
    link/ether 12:6e:dc:6c:6f:15 brd ff:ff:ff:ff:ff:ff
    inet 172.31.79.245/20 brd 172.31.79.255 scope global dynamic ens5
       valid_lft 2988sec preferred_lft 2988sec
    inet6 fe80::106e:dcff:fe6c:6f15/64 scope link
       valid_lft forever preferred_lft forever
3: fan-252: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue state UP group default qlen 1000
    link/ether de:1d:bf:6a:7a:8b brd ff:ff:ff:ff:ff:ff
    inet 252.79.245.1/8 scope global fan-252
       valid_lft forever preferred_lft forever
    inet6 fe80::dc1d:bfff:fe6a:7a8b/64 scope link
       valid_lft forever preferred_lft forever
4: ftun0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 8951 qdisc noqueue master fan-252 state UNKNOWN group default qlen 1000
    link/ether de:1d:bf:6a:7a:8b brd ff:ff:ff:ff:ff:ff
    inet6 fe80::dc1d:bfff:fe6a:7a8b/64 scope link
       valid_lft forever preferred_lft forever
5: flannel.1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 8951 qdisc noqueue state UNKNOWN group default
    link/ether 8e:ee:02:5a:3f:6a brd ff:ff:ff:ff:ff:ff
    inet 10.1.60.0/32 scope global flannel.1
       valid_lft forever preferred_lft forever
    inet6 fe80::8cee:2ff:fe5a:3f6a/64 scope link
       valid_lft forever preferred_lft forever
6: cni0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 8951 qdisc noqueue state UP group default qlen 1000
    link/ether 72:e3:16:d9:8f:30 brd ff:ff:ff:ff:ff:ff
    inet 10.1.60.1/24 scope global cni0
       valid_lft forever preferred_lft forever
    inet6 fe80::70e3:16ff:fed9:8f30/64 scope link
       valid_lft forever preferred_lft forever
8: veth229251aa@if3: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 8951 qdisc noqueue master cni0 state UP group default
    link/ether 9e:97:b0:54:b7:41 brd ff:ff:ff:ff:ff:ff link-netns cni-c976d5f5-1022-6723-2a3d-b8d5b3cb680c
    inet6 fe80::9c97:b0ff:fe54:b741/64 scope link
       valid_lft forever preferred_lft forever
-----

Now do the same thing on GCE, note only fan here:
-----
$ juju run --unit flannel/0 'network-get cni'
bind-addresses:
- macaddress: 8a:1c:c2:fc:4c:b3
  interfacename: fan-252
  addresses:
  - hostname: ""
    address: 252.0.87.1
    cidr: 252.0.0.0/8
egress-subnets:
- 252.0.87.1/32
ingress-addresses:
- 252.0.87.1
-----

With the following interfaces being available on flannel/0:
-----
$ ip a
1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
    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
    inet6 ::1/128 scope host
       valid_lft forever preferred_lft forever
2: ens4: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1460 qdisc mq state UP group default qlen 1000
    link/ether 42:01:0a:f0:00:57 brd ff:ff:ff:ff:ff:ff
    inet 10.240.0.87/32 scope global dynamic ens4
       valid_lft 59588sec preferred_lft 59588sec
    inet6 fe80::4001:aff:fef0:57/64 scope link
       valid_lft forever preferred_lft forever
3: fan-252: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1450 qdisc noqueue state UP group default qlen 1000
    link/ether 8a:1c:c2:fc:4c:b3 brd ff:ff:ff:ff:ff:ff
    inet 252.0.87.1/8 scope global fan-252
       valid_lft forever preferred_lft forever
    inet6 fe80::881c:c2ff:fefc:4cb3/64 scope link
       valid_lft forever preferred_lft forever
4: ftun0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1410 qdisc noqueue master fan-252 state UNKNOWN group default qlen 1000
    link/ether 8a:1c:c2:fc:4c:b3 brd ff:ff:ff:ff:ff:ff
    inet6 fe80::881c:c2ff:fefc:4cb3/64 scope link
       valid_lft forever preferred_lft forever
5: flannel.1: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1410 qdisc noqueue state UNKNOWN group default
    link/ether a6:ca:a4:e5:2c:27 brd ff:ff:ff:ff:ff:ff
    inet 10.1.19.0/32 scope global flannel.1
       valid_lft forever preferred_lft forever
    inet6 fe80::a4ca:a4ff:fee5:2c27/64 scope link
       valid_lft forever preferred_lft forever
-----

One big difference that I see is AWS has ensX with a 172.31.x.x addr, while GCE has ensX with a 10.240.x.y. Is there something in juju that filters 10.x and not 172.x such that 10.x interfaces are precluded from something like 'network-get'?