A common analysis is counting values, grouped on common attribute. e.g. What are the counts of passing, failing, missing tests.

IN SQL, we can do this:

SELECT name, status, COUNT(*) 
FROM tests 
WHERE date > NOW() - INTERVAL 30 DAY
AND name IN ('test1', 'test2')
GROUP BY name, status

And you might expect to see:

name	status	COUNT(*)
test1	passing	100
test1	failed	10
test1	aborted	30
test2	passing	100

To perform aggregations in elasticsearch:

curl https://example.com:1337/tests_index/tset/_search -d @blahblah.json

where blahblah.json looks like:

{
  "size": 0,
  "aggs": {
    "test_name" : {
      "terms" : { "field" : "name", "size": 100 },
      "aggs" : {
        "status" : {
          "terms" : { "field": "status"}
        }
      }
    }
  },
  "query": {
    "bool" : {
      "filter" : [
        { 
          "range" : {
            "time.start" : {
              "gte": "now-1M"
            }
          }
        },
        {
          "terms": {
            "name": [
              "test1",
              "test2"
            ]
          }
        }
      ]
    }
  }
}

The query has a bool, which is where we put our filters. In my case, I don't need scoring. filter takes a list of AND conditions. I've done a range check to look for results in the last month and also limited results to those matching my test name.

The aggregation is a terms which is a value count. The top-level aggregration is done on the test name, equivalent to the first column in the SQL group by. Then sub-aggregation is on the status term. Sub-aggregations can be nested to continue the "select" + "group by".