Adds usage data for ILM

[colings86]

Currently the response from the x-pack usage API would be:

"ilm": {
    "policy_count": 4,
    "policy_stats": [
      {
        "phases": {
          "warm": {
            "after": "0ms",
            "actions": [
              "allocate",
              "forcemerge"
            ]
          },
          "hot": {
            "after": "0ms",
            "actions": [
              "rollover"
            ]
          },
          "delete": {
            "after": "2m",
            "actions": [
              "delete"
            ]
          }
        }
      },
      {
        "phases": {
          "hot": {
            "after": "0ms",
            "actions": [
              "rollover"
            ]
          },
          "delete": {
            "after": "2m",
            "actions": [
              "delete"
            ]
          }
        }
      },
      {
        "phases": {
          "warm": {
            "after": "0ms",
            "actions": [
              "allocate",
              "shrink",
              "forcemerge"
            ]
          },
          "cold": {
            "after": "1m",
            "actions": [
              "allocate"
            ]
          },
          "hot": {
            "after": "0ms",
            "actions": [
              "rollover"
            ]
          },
          "delete": {
            "after": "2m",
            "actions": [
              "delete"
            ]
          }
        }
      },
      {
        "phases": {
          "warm": {
            "after": "0ms",
            "actions": [
              "allocate",
              "forcemerge"
            ]
          },
          "cold": {
            "after": "1m",
            "actions": [
              "allocate"
            ]
          },
          "hot": {
            "after": "0ms",
            "actions": [
              "rollover"
            ]
          },
          "delete": {
            "after": "2m",
            "actions": [
              "delete"
            ]
          }
        }
      }
    ]
  },

[elasticmachine]

Pinging @elastic/es-core-infra

[colings86]

@elasticmachine test this please

[colings86]

@elasticmachine test this please

[talevy]

Do we know how people are expected to use this data? I believe it does a great job providing all the policy usage info on a per-policy basis for users to group and keep track of things.

What I am not clear on is, who is expected to consume this data? is it other Elasticsearch toolings, or users?
For example, if counts of action usages are important, maybe this structure would pre-calculate things for the user:

{
 <phase_name>: {
     "after": ["1m", "5d", "13d"],
     "actions": {
        <action_name>: <num_times_action_used_in_this_phase>,
        ...
     }
 }
}

@eskibars, since Colin is out, I would love your input on this since I am missing some context. thanks!

[talevy]

After speaking with @eskibars in a different channel, I learned that these documents are intended to be aggregated as documents in ES, and so their structure should reflect that intention.

If that is the case, should the phase be contained in a field value that can be used in term aggregations?

[talevy]

I have left a few comments

[talevy]

after is now minimum_age

[talevy]

I think it makes sense to store this in milliseconds so that we can do statistics on it?

[talevy]

at first I thought this format would make it difficult to know how many policies didn't use a certain field, but after a few test queries, it looks like this is sufficient

[colings86]

@talevy I addressed your comment, could you take another look?

[talevy]

I think this will be good!

should we add yaml tests?