Fail with error when autodiscover providers have no defined configs

[andrewvc]

If one accidentally misconfigures an autodiscover provider and there are no defined configs no error is raised today. This patch adds an error in this situation and also does a better job adding context to other initialization errors.

You can test this with the following config which looks like it should work in heartbeat:

heartbeat.autodiscover:
  providers:
    - type: docker
      templates:
          condition:
            contains:
              docker.container.image: redis
          config:
            - type: tcp
              hosts: ["${data.host}:${data.port}"]
              schedule: "@every 1s"
              timeout: 1s

However, it does not because templates should be an array. The correct config is below:

heartbeat.autodiscover:
  providers:
    - type: docker
      templates:
        - condition:
            contains:
              docker.container.image: redis
          config:
            - type: tcp
              hosts: ["${data.host}:${data.port}"]
              schedule: "@every 1s"
              timeout: 1s

[exekias]

This is a valid config, for instance when people is using hints based autodiscover, it's ok to leave templates empty:

filebeat.autodiscover:
  providers:
    - type: kubernetes
      hints.enabled: true

We would need to account for that for an error like this

[exekias]

Another check we could do here is to ensure len(c.Configs) > 0. I think in most cases this is were bad configs happen? when the user does:

autodiscover:
  providers:
    type: kubernetes
    templates:
      - equals:
          container.image: ...
        config:
          type: container

But config is expected to be a list, so a - is missing

[andrewvc]

I'm a bit confused, is this section used when hints are enabled?

[exekias]

Thank you for improving this!

[andrewvc]

@exekias I just pushed a fix that takes hints into account for providers that support them.

[exekias]

sorry, I approved as the change looks good to me, but it seems one test is failing, we will probably need to adjust it after this change. A changelog will also be needed

[Randy-312]

Kudos on this one!... it would've saved us a lot of time when we had our configuration mis-done.

[andrewvc]

Sorry for the delay here, hoping to come back to this one within the next few weeks, I'm sidetracked on other PRs ATM