spec_helper_acceptance: add abbility to enable Ipv6 in beaker hosts

[b4ldr]

Enabled by default
Also see: voxpupuli/modulesync_config#694

[ekohl]

Does this work in Docker?

[b4ldr]

Yes its been in use in the unbound module for some time and works as expected. I have also used it in other projects but that was some time ago

[b4ldr]

i originally used sysctl net.ipv6.conf.all.disable_ipv6=0 but sysctl isn't installed on fedora by default.

[b4ldr]

Does this work in Docker?

Tested and working however when i tested locally (with a plain docker image) i noticed that /proc/sys is mounted readonly. As such im gussing beaker takes care of making sure this is mounted with rw?

[b4ldr]

i noticed that /proc/sys is mounted readonly. As such im gussing beaker takes care of making sure this is mounted with rw?

confirmed beaker starts the container with --privileged which dosn't have this issue

[ekohl]

I think that does mean you modify sysctl from the host, which can be nasty as well. I'll submit a PR to unbound to use some ULA subnet instead of a global one to see if that passes.

[b4ldr]

I think that does mean you modify sysctl from the host, which can be nasty as well.

that's not correct for this parameter and most kernel parameters they have there own cgroup/namespace see below:

~/git/puppet $ cat /proc/sys/net/ipv6/conf/all/disable_ipv6                            
0
~/git/puppet $ docker run --privileged --rm -it fedora:32 /bin/bash                       
[root@98a76107e9d8 /]# cat /proc/sys/net/ipv6/conf/all/disable_ipv6 
1

I'll submit a PR to unbound to use some ULA subnet instead of a global one to see if that passes.

hmm setting the following "fixed-cidr-v6": "fd00::/8" in docker.yaml may work as well however if one really just needs the ::1 loopback which i suspect is the case in most situations then i think i rather lean towards this solution

*fyi which ever fix needs to be made in modsync_config not puppet-unbound

[ekohl]

I thought a bit more about it. One thing I like about doing it via a sysctl is that you are less dependent on the environment so you can reproduce it locally. Later I'll see about merging your PRs but now it's time to call it a day.

[ekohl]

I forgot about this, but I've since taken a difference approach and use RSpec.add_setting instead of adding parameters to configure_beaker to tune behavior. I also pushed the setup_acceptance_node.pp pattern into the module which makes things easier to use. That means it now needs a rebase.

I'd also be ok with not solving it in this library but rather keep it up to the modules to fix it.

[b4ldr]

Sorry i missed this, have rebased now

I'd also be ok with not solving it in this library but rather keep it up to the modules to fix it.

On the one hand this seems to only affect a small subset of modules so perhaps not worth it. On the other hand its a very minor addition and could save someone from scratching there head when implementing acceptance test. as such i would favour adding it