This message was deleted.

You could write another plugin to do the conversion. You could do something like this where

my_transform_plugin

gets the value from another plugin and converts it:

yaml
config:
  ssh:
    proxyjump:
      _plugin: my_transform_plugin
      value:
        _plugin: aws_inventory
        [...]

Ah cool, feeding plugins to plugins

You could also abstract away the

aws_inventory

bits by having a plugin that, when invoked, returns a

_plugin: aws_inventory

block

So will it iteratively evaluate until idempotent? Or what you're saying is that it won't?

Yeah it evaluates bottom-up and will continue to evaluate until the value no longer contains

_plugin

Okay I got my plugin written and it seems to be working, however when I run bolt I get an odd error. To simplify the config I removed all plugins and run this:

yaml
---
groups:
  - name: aws
    groups:
      - name: aws_dev
        groups:
          - name: aws_dev_buster
            targets:
            - 10.225.0.216
            config:
              ssh:
                proxyjump: "admin@34.219.87.226"

And the error I'm getting is:

stdout: '{ "items": [\n' +
    '{"target":"10.225.0.216","action":"script","object":null,"status":"failure","value":{"_error":{"details":{},"kind":"puppetlabs.tasks/connect-error","msg":"Failed to connect to 10.225.0.216: No such file or directory - ssh: ssh -l admin -W 10.225.0.216:22 34.219.87.226","issue_code":"CONNECT_ERROR"}}}\n' +
    '],\n' +
    '"target_count": 1, "elapsed_time": 0 }\n',
  stderr: '\x1B[33mCLI arguments ["host-key-check", "run-as"] might be overridden by Inventory: /deploy/inventory.yaml [ID: cli_overrides]\x1B[0m\n' +
    '\x1B[0m'

This seems to indicate that the ssh.config.proxyjump value is getting parsed in the wrong way, tho I'm not sure if it's bolt or ruby's

net/ssh/proxy/jump

doing the incorrect parsing :\