add index metrics via alias

[matsumana]

Please review and merge #85 first.

I would like to monitoring whether alias settings are correct.
I realize it to monitoring index metrics via alias.

so I added the metrics below:

# HELP elasticsearch_aliases_docs_primary Count of documents which only primary shards via alias
# TYPE elasticsearch_aliases_docs_primary gauge
elasticsearch_aliases_docs_primary{alias="foo",cluster="test",index="foo_1"} 2
elasticsearch_aliases_docs_primary{alias="fooA",cluster="test",index="foo_1"} 2
elasticsearch_aliases_docs_primary{alias="fooB",cluster="test",index="foo_2"} 3

[dominikschulz]

LGTM

[matsumana]

Hello, Please also review this PR.

[zwopir]

can you please add tests fixtures for the versions != 5.x.x and verify they look similar to yours. If they don't, we need to think about a switch to enable indices stats metrics

[zwopir]

can you please add fixtures for 1.7.6 and 2.4.5 (compare nodes_test.go)?

[matsumana]

@zwopir
I added fixtures for 1.7.6, 2.4.5 and 5.4.2
Would you please review again?

[zwopir]

thanks for the update. I'm not really happy with the PR:

• the collector does two API calls, one of them is redundant - at least if you restructure the code. I haven't yet thought how to refactor
• within the alias collector you use results from the _aliasAPI call as map key for the results of the _all/_stats call. This should be consistent ES wise, but I wouldn't rely on it.

Could you elaborate a bit more on the use case of this PR? I would like to keep the exporter and especially the collectors as well structured and uncoupled as possible

[matsumana]

For example, There are the following index and alias.

• foo_1 (index)
• foo (alias)

An application refers to the index via alias.
However there is a possibility that it can not be referred due to a program bug or misoperation.
so I want to monitoring whether alias settings are correct.

I think I can realize it, by this metric(elasticsearch_aliases_docs_primary) and following PromQL.

absent(elasticsearch_aliases_docs_primary{alias="foo"}) == 1

[zwopir]

I understand your use case, but I don't think its a use case that should be monitored via metrics. Misconfiguration is generally speaking not a thing you want to monitor constantly (i.e. with a scape interval in the range of seconds), but just after a change within the ES cluster or after a software release that changes the index/alias settings. Especially if it is done with a "nagios style" ok/not ok metric. So it's more a matter of release management and QA, not of operations.
I would therefore reject this PR.

What do you think, @dominikschulz and @falschparker82 (speaking a bit more from the ES side)?

@zwopir raised some valid concerns that should be resolved first

[matsumana]

I understand and agree with you.
I will consider other way.