fix _/node/stats related metrics

[0x46616c6b]

Since Elasticsearch 7.13 the node related metrics are broken. It relates to a API change in Elasticsearch which breaks the unmarshling:

level=warn ts=2021-06-15T09:53:13.626727Z caller=nodes.go:1871 msg="failed to fetch and decode node stats" err="json: cannot unmarshal array into Go struct field NodeStatsNodeResponse.Nodes.http of type int"

The HTTP field will only checked for the roles and also only as len(HTTP), so it seems it is not important to have the type here.

Example for http field in the response (Elasticsearch 7.13)

    {
      "id": 299723549,
      "agent": "curl/7.61.1",
      "local_address": "127.0.0.1:9200",
      "remote_address": "127.0.0.1:41480",
      "last_uri": "/",
      "opened_time_millis": 1623753253078,
      "closed_time_millis": 1623753253078,
      "last_request_time_millis": 1623753253078,
      "request_count": 1,
      "request_size_bytes": 0
    },

Related #419

[sysadmind]

There is a test that is skipped here for 7.13. This will need to be removed as well to ensure that nothing else is broken.

[Skunnyk]

Thank you ! Will this works on ES < 7.13 ?

[0x46616c6b]

Thank you ! Will this works on ES < 7.13 ?

Yes, the automated tests and a manual test on a 6.8 cluster works.

[adammike]

@sysadmind is there a ci pipline that automatically deploys this image someplace? or does it have to wait for a release?

[sysadmind]

The available docker images are here: https://quay.io/repository/prometheuscommunity/elasticsearch-exporter?tab=tags
There is a master tag if you like the bleeding edge 😄

[adammike]

heh, can't get any more broken than it is now.. for 7.13 anyway.

[adammike]

Deployed the master image for two of our es clusters, works great 👍

[gkulkarni32]

@sysadmind Wanted to know by when the changes will be available in the exporter image (1.2.1?)

[sysadmind]

v1.2.1 has been released

[gkulkarni32]

Thank you @sysadmind , will test it out.