Troubleshooting searches

When you query your data, Elasticsearch may return an error, no search results, or results in an unexpected order. This guide describes how to troubleshoot searches.

Elasticsearch returns an index_not_found_exception when the data stream, index or alias you try to query does not exist. This can happen when you misspell the name or when the data has been indexed to a different data stream or index.

Use the exists API to check whether a data stream, index, or alias exists:

HEAD my-data-stream

Use the data stream stats API to list all data streams:

response = client.indices.data_streams_stats(
  human: true
)
puts response

GET /_data_stream/_stats?human=true

Use the get index API to list all indices and their aliases:

response = client.indices.get(
  index: '_all',
  filter_path: '*.aliases'
)
puts response

GET _all?filter_path=*.aliases

Instead of an error, it is possible to retrieve partial search results if some of the indices you’re querying are unavailable. Set ignore_unavailable to true:

response = client.search(
  index: 'my-alias',
  ignore_unavailable: true
)
puts response

GET /my-alias/_search?ignore_unavailable=true

When a search request returns no hits, the data stream or index may contain no data. This can happen when there is a data ingestion issue. For example, the data may have been indexed to a data stream or index with another name.

Use the count API to retrieve the number of documents in a data stream or index. Check that count in the response is not 0.

response = client.count(
  index: 'my-index-000001'
)
puts response

GET /my-index-000001/_count

Querying a field that does not exist will not return any results. Use the field capabilities API to check whether a field exists:

response = client.field_caps(
  index: 'my-index-000001',
  fields: 'my-field'
)
puts response

GET /my-index-000001/_field_caps?fields=my-field

If the field does not exist, check the data ingestion process. The field may have a different name.

If the field exists, the request will return the field’s type and whether it is searchable and aggregatable.

{
  "indices": [
    "my-index-000001"
  ],
  "fields": {
    "my-field": {
      "keyword": {
        "type": "keyword",         
        "metadata_field": false,
        "searchable": true,        
        "aggregatable": true       
      }
    }
  }
}

A field’s capabilities are determined by its mapping. To retrieve the mapping, use the get mapping API:

response = client.indices.get_mapping(
  index: 'my-index-000001'
)
puts response

GET /my-index-000001/_mappings

If you query a text field, pay attention to the analyzer that may have been configured. You can use the analyze API to check how a field’s analyzer processes values and query terms:

response = client.indices.analyze(
  index: 'my-index-000001',
  body: {
    field: 'my-field',
    text: 'this is a test'
  }
)
puts response

GET /my-index-000001/_analyze
{
  "field" : "my-field",
  "text" : "this is a test"
}

To change the mapping of an existing field, refer to Changing the mapping of a field.

Use the exists query to check whether there are documents that return a value for a field. Check that count in the response is not 0.

response = client.count(
  index: 'my-index-000001',
  body: {
    query: {
      exists: {
        field: 'my-field'
      }
    }
  }
)
puts response

GET /my-index-000001/_count
{
  "query": {
    "exists": {
      "field": "my-field"
    }
  }
}

If the field is aggregatable, you can use aggregations to check the field’s values. For keyword fields, you can use a terms aggregation to retrieve the field’s most common values:

response = client.search(
  index: 'my-index-000001',
  filter_path: 'aggregations',
  body: {
    size: 0,
    aggregations: {
      top_values: {
        terms: {
          field: 'my-field',
          size: 10
        }
      }
    }
  }
)
puts response

GET /my-index-000001/_search?filter_path=aggregations
{
  "size": 0,
  "aggs": {
    "top_values": {
      "terms": {
        "field": "my-field",
        "size": 10
      }
    }
  }
}

For numeric fields, you can use the stats aggregation to get an idea of the field’s value distribution:

response = client.search(
  index: 'my-index-000001',
  filter_path: 'aggregations',
  body: {
    aggregations: {
      "my-num-field-stats": {
        stats: {
          field: 'my-num-field'
        }
      }
    }
  }
)
puts response

GET my-index-000001/_search?filter_path=aggregations
{
  "aggs": {
    "my-num-field-stats": {
      "stats": {
        "field": "my-num-field"
      }
    }
  }
}

If the field does not return any values, check the data ingestion process. The field may have a different name.

For time-series data, confirm there is non-filtered data within the attempted time range. For example, if you are trying to query the latest data for the @timestamp field, run the following to see if the max @timestamp falls within the attempted range:

response = client.search(
  index: 'my-index-000001',
  sort: '@timestamp:desc',
  size: 1
)
puts response

GET my-index-000001/_search?sort=@timestamp:desc&size=1

When a query returns unexpected results, Elasticsearch offers several tools to investigate why.

The validate API enables you to validate a query. Use the rewrite parameter to return the Lucene query an Elasticsearch query is rewritten into:

response = client.indices.validate_query(
  index: 'my-index-000001',
  rewrite: true,
  body: {
    query: {
      match: {
        "user.id": {
          query: 'kimchy',
          fuzziness: 'auto'
        }
      }
    }
  }
)
puts response

GET /my-index-000001/_validate/query?rewrite=true
{
  "query": {
    "match": {
      "user.id": {
        "query": "kimchy",
        "fuzziness": "auto"
      }
    }
  }
}

Use the explain API to find out why a specific document matches or doesn’t match a query:

response = client.explain(
  index: 'my-index-000001',
  id: 0,
  body: {
    query: {
      match: {
        message: 'elasticsearch'
      }
    }
  }
)
puts response

GET /my-index-000001/_explain/0
{
  "query" : {
    "match" : { "message" : "elasticsearch" }
  }
}

The profile API provides detailed timing information about a search request. For a visual representation of the results, use the Search Profiler in Kibana.

Index settings can influence search results. For example, the index.query.default_field setting, which determines the field that is queried when a query specifies no explicit field. Use the get index settings API to retrieve the settings for an index:

response = client.indices.get_settings(
  index: 'my-index-000001'
)
puts response

GET /my-index-000001/_settings

You can update dynamic index settings with the update index settings API. Changing dynamic index settings for a data stream requires changing the index template used by the data stream.

For static settings, you need to create a new index with the correct settings. Next, you can reindex the data into that index. For data streams, refer to Change a static index setting for a data stream.

Slow logs can help pinpoint slow performing search requests. Enabling audit logging on top can help determine query source. Add the following settings to the elasticsearch.yml configuration file to trace queries. The resulting logging is verbose, so disable these settings when not troubleshooting.

xpack.security.audit.enabled: true
xpack.security.audit.logfile.events.include: _all
xpack.security.audit.logfile.events.emit_request_body: true

Refer to Advanced tuning: finding and fixing slow Elasticsearch queries for more information.