[DOC] Describe differences between Ansible group variables and group/node data

[ipspace]

Describe the bug

When setting an Ansible variable in a custom group, Ansible inventory correctly inherits the changed setting while the adjust_inventory_host does not.

To Reproduce

See #1007 for details.

Expected behavior

ansible-inventory and netlab connect should get the same idea what ansible_connection variable value should be.

Lab topology

provider: clab

groups:
  linux_ssh:
    members: [ h ]
    vars:
      ansible_connection: ssh

nodes:
  h:
    device: linux

Version

All netlab versions

Additional context

There are two ways to solve this challenge:

1. Consider custom groups and their vars settings in adjust_inventory_host
2. Do full inheritance before storing the data in the snapshot file

(1) might seem more complex, but we'd have to have the code to do the inheritance somewhere anyway, so maybe it doesn't matter that much. (2) would result in extra data in netlab inspect or netlab create --output yaml.

[sdktr]

This comment should probably have gone here:
#1007 (comment)

[ipspace]

After mulling this over a bit more: Ansible group variables within custom groups seems a fringe functionality, and I have no idea whether anyone is using them. It doesn't seem worth it to add another bit of potentially-brittle code to support that, so here's plan B: document what's going on and what a better way of handling the custom attributes is in https://netlab.tools/groups/#group-variables

Any thoughts/objections @sdktr @ssasso @jbemmel @petercrocker

[sdktr]

If you mean: let's not output ansible-groups anymore, I'm all for that. netlab has extensive logic to generate host specific variables. These can be put into the applicably ansible host vars directly.

Why do we need ansible groups (custom or default ones) anyway given that we already know the host specific vars?

[ipspace]

Why do we need ansible groups (custom or default ones) anyway given that we already know the host specific vars?

There are a few reasons I decided to use Ansible groups when designing netlab:

• There are fewer variables per host (I guess that doesn't matter any longer considering how big the data model is)
• You can change Ansible-related stuff in one place (group variable) without recreating Ansible inventory from the lab topology
• You can add any host- or group variables you wish -- that's why I'm using host/group directories instead of simple YAML files
• You can add stuff to Ansible groups that can later be used in templates without going through the netlab attribute definition process.

Unfortunately, I have no idea how people use netlab, but considering how creative networking engineers tend to get while trying to hammer a star-shaped peg into triangular hole, I wouldn't be surprised if someone built an ingenious solution relying on the above.

Long story short: I'll document the shortcomings and the differences between Ansible variables and netlab connect, and point to the new change ansible connection method example.