Docker minikube network not automatically removed at minikube stop or start

[srilumpa]

Hi,

I upgraded this morning my minikube cluster from version 1.14.0 to 1.16.0 and it seems the docker network usage is not fully handled. I was unable to restart my minikube cluster without manually deleting the docker network created by minikube.

Steps to reproduce the issue:

0. Having a running minikube clusterwith minikube in version 1.14.0
1. minikube stop && minikube start

Full output of failed command:

✋  Stopping node "minikube"  ...
🛑  Powering off "minikube" via SSH ...
🛑  1 nodes stopped.
😄  minikube v1.16.0 on Ubuntu 20.04
🆕  Kubernetes 1.20.0 is now available. If you would like to upgrade, specify: --kubernetes-version=v1.20.0
✨  Using the docker driver based on existing profile
👍  Starting control plane node minikube in cluster minikube
🔄  Restarting existing docker container for "minikube" ...
🐳  Preparing Kubernetes v1.18.3 on Docker 19.03.8 ...| E1230 09:47:23.082385   13400 start.go:99] Unable to get host IP: network inspect: error parsing network inspect output: "{\"Name\": \"minikube\",\"Driver\": \"bridge\",\"Subnet\": \"192.168.49.0/24\",\"Gateway\": \"192.168.49.1\",\"MTU\": , \"ContainerIPs\": [\"192.168.49.2/24\"]}\n"


❌  Exiting due to GUEST_START: Failed to setup kubeconfig: network inspect: error parsing network inspect output: "{\"Name\": \"minikube\",\"Driver\": \"bridge\",\"Subnet\": \"192.168.49.0/24\",\"Gateway\": \"192.168.49.1\",\"MTU\": , \"ContainerIPs\": [\"192.168.49.2/24\"]}\n"

😿  If the above advice does not help, please let us know: 
👉  https://github.com/kubernetes/minikube/issues/new/choose

At that step, I simply deleted my cluster and recreated it:

1. minikube delete && minikube start --cpus=4 --memory=8g --disk-size=40g

And it failed outputing the following:

😄  minikube v1.16.0 on Ubuntu 20.04
✨  Automatically selected the docker driver. Other choices: virtualbox, none
👍  Starting control plane node minikube in cluster minikube
🚜  Pulling base image ...
💾  Downloading Kubernetes v1.20.0 preload ...
    > preloaded-images-k8s-v8-v1....: 491.00 MiB / 491.00 MiB  100.00% 25.00 Mi
🔥  Creating docker container (CPUs=4, Memory=8192MB) ...
E1230 09:49:23.788509   15496 network_create.go:81] error while trying to create network create network minikube 192.168.49.0/24: docker network create --driver=bridge --subnet=192.168.49.0/24 --gateway=192.168.49.1 -o --ip-masq -o --icc -o com.docker.network.driver.mtu=1500 --label=created_by.minikube.sigs.k8s.io=true minikube: exit status 1
stdout:

stderr:
Error response from daemon: network with name minikube already exists
❗  Unable to create dedicated network, this might result in cluster IP change after restart: un-retryable: create network minikube 192.168.49.0/24: docker network create --driver=bridge --subnet=192.168.49.0/24 --gateway=192.168.49.1 -o --ip-masq -o --icc -o com.docker.network.driver.mtu=1500 --label=created_by.minikube.sigs.k8s.io=true minikube: exit status 1
stdout:

stderr:
Error response from daemon: network with name minikube already exists

🐳  Preparing Kubernetes v1.20.0 on Docker 20.10.0 ...| E1230 09:49:33.959739   15496 start.go:99] Unable to get host IP: network inspect: error parsing network inspect output: "{\"Name\": \"minikube\",\"Driver\": \"bridge\",\"Subnet\": \"192.168.49.0/24\",\"Gateway\": \"192.168.49.1\",\"MTU\": , \"ContainerIPs\": []}\n"


❌  Exiting due to GUEST_START: Failed to setup kubeconfig: network inspect: error parsing network inspect output: "{\"Name\": \"minikube\",\"Driver\": \"bridge\",\"Subnet\": \"192.168.49.0/24\",\"Gateway\": \"192.168.49.1\",\"MTU\": , \"ContainerIPs\": []}\n"

😿  If the above advice does not help, please let us know: 
👉  https://github.com/kubernetes/minikube/issues/new/choose

I looked into my docker networks and found the minikube network still existing. I removed it and was able to successfully start my new cluster:

1. docker network rm minikube
2. minikube start

😄  minikube v1.16.0 on Ubuntu 20.04
✨  Using the docker driver based on existing profile
👍  Starting control plane node minikube in cluster minikube
🏃  Updating the running docker "minikube" container ...
🐳  Preparing Kubernetes v1.20.0 on Docker 20.10.0 ...
    ▪ Generating certificates and keys ...
    ▪ Booting up control plane ...
    ▪ Configuring RBAC rules ...
🔎  Verifying Kubernetes components...
🌟  Enabled addons: storage-provisioner, default-storageclass
🏄  Done! kubectl is now configured to use "minikube" cluster and "default" namespace by default

I don't know if it is only linked to the upgrade and if it is still relevant with future upgrades but, in doubt, I wanted to report this issue.

[afbjorklund]

This is a side-effect of #10033, it is trying to inspect the old network - and then fails to parse the results.

Your workaround is the suggested one.

[medyagh]

I belive this was fixed by this PR #10053
yet deleting minikube and starting it agian with old version should still work.

please feel free to open if u have more questions