ignore_docker_version is ignored

[yankcrime]

The ignore_docker_version option appears to be ignored. However, passing --ignore-docker-version works as expected.

RKE version:

rke version v1.2.3

Docker version: (docker version,docker info preferred)

20.10.0 (CE)

Operating system and kernel: (cat /etc/os-release, uname -r preferred)

Ubuntu 20.04

Type/provider of hosts: (VirtualBox/Bare-metal/AWS/GCE/DO)

VMware ESXi

cluster.yml file:

cluster_name: vspherenetes
ssh_agent_auth: true
ignore_docker_version: true
kubernetes_version: "v1.19.4-rancher1-1"

nodes:
    - address: 192.168.1.92
      user: nick
      role:
        - controlplane
        - etcd
    - address: 192.168.1.12
      user: nick
      role:
        - worker
    - address: 192.168.1.27
      user: nick
      role:
        - worker
    - address: 192.168.1.96
      user: nick
      role:
        - worker

Steps to Reproduce:
rke up

Results:

FATA[0000] Unsupported Docker version found [20.10.0] on host [192.168.1.67], supported versions are [1.13.x 17.03.x 17.06.x 17.09.x 18.06.x 18.09.x 19.03.x]
zsh: exit 1     rke up

Retrying with the --ignore-docker-version set results in a successful cluster deployment.

[cesarandreslopez]

Noticed the same issue here

Retrying with the --ignore-docker-version set results in a successful cluster deployment.

[TemaSM]

Confirmed
rke: v1.2.4-rc9
Docker: 20.10.2, build 2291f61
system: Ubuntu-2004-focal-64-minimal 5.4.0-47-generic

[dmpe]

Also confirmed

[dekimpew]

Confirmed for version 1.2.5.
rke: v1.2.5
Docker: 20.10.3, build 48d30b5
OS: CentOS Linux release 7.9.2009 (Core)

[Minizinger]

Still a problem with rke version v1.3.13 (works with --ignore-docker-version flag)

[tashima42]

After a brief look at the code, I think I may have found the problem, still needs more investigation.

For this scenario, I'm assuming ignore_docker_version: true in cluster.yaml and that the flag --ignore-docker-version isn't provided.

The config is assigned here, so rkeConfig.IgnoreDockerVersion is set to true

rke/cmd/up.go

Line 329 in 5aaec0c

rkeConfig, err = cluster.ParseConfig(clusterFile)

Right after, it replaces the value of rkeConfig.IgnoreDockerVersion with a pointer to the value of the CLI flag ignore-docker-version

rke/cmd/up.go

Lines 336 to 337 in 5aaec0c

	ignoreDockerVersion := ctx.Bool("ignore-docker-version")
	rkeConfig.IgnoreDockerVersion = &ignoreDockerVersion

The issue is that, by default, if no value was provided, ctx.Bool returns false, so the value will always be false. This can be verified in the code for the release v1.22.5 for the package urfave/cli
https://github.com/urfave/cli/blob/850de854cda08e97afb500b72967f1a56f0330e5/flag_bool.go#L54-L56
https://github.com/urfave/cli/blob/850de854cda08e97afb500b72967f1a56f0330e5/flag_bool.go#L99-L109

A possible solution would be to check if the flag was actually set and only replace the value in the config if it was

if ctx.IsSet("ignore-docker-version") {
  ignoreDockerVersion := ctx.Bool("ignore-docker-version") 
  rkeConfig.IgnoreDockerVersion = &ignoreDockerVersion
}

I'm not sure about the solution, so further investigation is needed, but this can be used as a starting point.