Juju causes race condition for flannel when deploying charms onto both host and containers on the same host

[Andrey Grebennikov]

Juju 2.6.5, deploying CDK (Flannel in particular)
Originally the baremetal node is deployed via MAAS with the interface "bondm" and the IP address is assigned by MAAS to it.
Later, once some units are assigned to run in the containers on the same node, Juju creates the bridge "br-bondm", assigns the same IP to it and attaches containres' interfaces to this bridge.
However, binding information isn't being updated on the unit, but rather new info is appended:

ubuntu@infra1:~/k8s-vxflex$ juju run --unit flannel/8 'network-get cni'
bind-addresses:
- macaddress: 3c:fd:fe:b8:dd:10
  interfacename: bondm
  addresses:
  - hostname: ""
    address: 100.82.38.180
    cidr: 100.82.38.0/24
- macaddress: 3c:fd:fe:b8:dd:10
  interfacename: br-bondm
  addresses:
  - hostname: ""
    address: 100.82.38.180
    cidr: 100.82.38.0/24
egress-subnets:
- 100.82.38.180/32
ingress-addresses:
- 100.82.38.180
- 100.82.38.180

As Flannel receives the info from the binding info and fulfills its systemd template, systemd script receives wrong interface name (bondm) and the service of Flannel can't start complaining "can't find IPv4 address on the interface bondm".

[Canonical Juju QA Bot (juju-qa-bot)]

This bug has not been updated in 2 years, so we're marking it Low importance. If you believe this is incorrect, please update the importance.