New

The executive guide to generative AI

Read more

Modify a data stream

edit

Change mappings and settings for a data stream

edit

Each data stream has a matching index template. Mappings and index settings from this template are applied to new backing indices created for the stream. This includes the stream’s first backing index, which is auto-generated when the stream is created.

Before creating a data stream, we recommend you carefully consider which mappings and settings to include in this template.

If you later need to change the mappings or settings for a data stream, you have a few options:

If your changes include modifications to existing field mappings or static index settings, a reindex is often required to apply the changes to a data stream’s backing indices. If you are already performing a reindex, you can use the same process to add new field mappings and change dynamic index settings. See Use reindex to change mappings or settings.

Add a new field mapping to a data stream

edit

To add a mapping for a new field to a data stream, following these steps:

  1. Update the index template used by the data stream. This ensures the new field mapping is added to future backing indices created for the stream.

    For example, my-data-stream-template is an existing index template used by my-data-stream.

    The following create or update index template request adds a mapping for a new field, message, to the template.

    PUT /_index_template/my-data-stream-template
    {
      "index_patterns": [ "my-data-stream*" ],
      "data_stream": { },
      "priority": 500,
      "template": {
        "mappings": {
          "properties": {
            "message": {                              
              "type": "text"
            }
          }
        }
      }
    }

    Adds a mapping for the new message field.

  2. Use the update mapping API to add the new field mapping to the data stream. By default, this adds the mapping to the stream’s existing backing indices, including the write index.

    The following update mapping API request adds the new message field mapping to my-data-stream.

    PUT /my-data-stream/_mapping
    {
      "properties": {
        "message": {
          "type": "text"
        }
      }
    }

    To add the mapping only to the stream’s write index, set the update mapping API’s write_index_only query parameter to true.

    The following update mapping request adds the new message field mapping only to my-data-stream's write index. The new field mapping is not added to the stream’s other backing indices.

    PUT /my-data-stream/_mapping?write_index_only=true
    {
      "properties": {
        "message": {
          "type": "text"
        }
      }
    }

Change an existing field mapping in a data stream

edit

The documentation for each mapping parameter indicates whether you can update it for an existing field using the update mapping API. To update these parameters for an existing field, follow these steps:

  1. Update the index template used by the data stream. This ensures the updated field mapping is added to future backing indices created for the stream.

    For example, my-data-stream-template is an existing index template used by my-data-stream.

    The following create or update index template request changes the argument for the host.ip field’s ignore_malformed mapping parameter to true.

    PUT /_index_template/my-data-stream-template
    {
      "index_patterns": [ "my-data-stream*" ],
      "data_stream": { },
      "priority": 500,
      "template": {
        "mappings": {
          "properties": {
            "host": {
              "properties": {
                "ip": {
                  "type": "ip",
                  "ignore_malformed": true            
                }
              }
            }
          }
        }
      }
    }

    Changes the host.ip field’s ignore_malformed value to true.

  2. Use the update mapping API to apply the mapping changes to the data stream. By default, this applies the changes to the stream’s existing backing indices, including the write index.

    The following update mapping API request targets my-data-stream. The request changes the argument for the host.ip field’s ignore_malformed mapping parameter to true.

    response = client.indices.put_mapping(
      index: 'my-data-stream',
      body: {
        properties: {
          host: {
            properties: {
              ip: {
                type: 'ip',
                ignore_malformed: true
              }
            }
          }
        }
      }
    )
    puts response
    PUT /my-data-stream/_mapping
    {
      "properties": {
        "host": {
          "properties": {
            "ip": {
              "type": "ip",
              "ignore_malformed": true
            }
          }
        }
      }
    }

    To apply the mapping changes only to the stream’s write index, set the put mapping API’s write_index_only query parameter to true.

    The following update mapping request changes the host.ip field’s mapping only for my-data-stream's write index. The change is not applied to the stream’s other backing indices.

    response = client.indices.put_mapping(
      index: 'my-data-stream',
      write_index_only: true,
      body: {
        properties: {
          host: {
            properties: {
              ip: {
                type: 'ip',
                ignore_malformed: true
              }
            }
          }
        }
      }
    )
    puts response
    PUT /my-data-stream/_mapping?write_index_only=true
    {
      "properties": {
        "host": {
          "properties": {
            "ip": {
              "type": "ip",
              "ignore_malformed": true
            }
          }
        }
      }
    }

Except for supported mapping parameters, we don’t recommend you change the mapping or field data type of existing fields, even in a data stream’s matching index template or its backing indices. Changing the mapping of an existing field could invalidate any data that’s already indexed.

If you need to change the mapping of an existing field, create a new data stream and reindex your data into it. See Use reindex to change mappings or settings.

Change a dynamic index setting for a data stream

edit

To change a dynamic index setting for a data stream, follow these steps:

  1. Update the index template used by the data stream. This ensures the setting is applied to future backing indices created for the stream.

    For example, my-data-stream-template is an existing index template used by my-data-stream.

    The following create or update index template request changes the template’s index.refresh_interval index setting to 30s (30 seconds).

    PUT /_index_template/my-data-stream-template
    {
      "index_patterns": [ "my-data-stream*" ],
      "data_stream": { },
      "priority": 500,
      "template": {
        "settings": {
          "index.refresh_interval": "30s"             
        }
      }
    }

    Changes the index.refresh_interval setting to 30s (30 seconds).

  2. Use the update index settings API to update the index setting for the data stream. By default, this applies the setting to the stream’s existing backing indices, including the write index.

    The following update index settings API request updates the index.refresh_interval setting for my-data-stream.

    response = client.indices.put_settings(
      index: 'my-data-stream',
      body: {
        index: {
          refresh_interval: '30s'
        }
      }
    )
    puts response
    PUT /my-data-stream/_settings
    {
      "index": {
        "refresh_interval": "30s"
      }
    }

To change the index.lifecycle.name setting, first use the remove policy API to remove the existing ILM policy. See Switch lifecycle policies.

Change a static index setting for a data stream

edit

Static index settings can only be set when a backing index is created. You cannot update static index settings using the update index settings API.

To apply a new static setting to future backing indices, update the index template used by the data stream. The setting is automatically applied to any backing index created after the update.

For example, my-data-stream-template is an existing index template used by my-data-stream.

The following create or update index template API requests adds new sort.field and sort.order index settings to the template.

PUT /_index_template/my-data-stream-template
{
  "index_patterns": [ "my-data-stream*" ],
  "data_stream": { },
  "priority": 500,
  "template": {
    "settings": {
      "sort.field": [ "@timestamp"],             
      "sort.order": [ "desc"]                    
    }
  }
}

Adds the sort.field index setting.

Adds the sort.order index setting.

If wanted, you can roll over the data stream to immediately apply the setting to the data stream’s write index. This affects any new data added to the stream after the rollover. However, it does not affect the data stream’s existing backing indices or existing data.

To apply static setting changes to existing backing indices, you must create a new data stream and reindex your data into it. See Use reindex to change mappings or settings.

Use reindex to change mappings or settings

edit

You can use a reindex to change the mappings or settings of a data stream. This is often required to change the data type of an existing field or update static index settings for backing indices.

To reindex a data stream, first create or update an index template so that it contains the wanted mapping or setting changes. You can then reindex the existing data stream into a new stream matching the template. This applies the mapping and setting changes in the template to each document and backing index added to the new data stream. These changes also affect any future backing index created by the new stream.

Follow these steps:

  1. Choose a name or index pattern for a new data stream. This new data stream will contain data from your existing stream.

    You can use the resolve index API to check if the name or pattern matches any existing indices, aliases, or data streams. If so, you should consider using another name or pattern.

    The following resolve index API request checks for any existing indices, aliases, or data streams that start with new-data-stream. If not, the new-data-stream* index pattern can be used to create a new data stream.

    response = client.indices.resolve_index(
      name: 'new-data-stream*'
    )
    puts response
    GET /_resolve/index/new-data-stream*

    The API returns the following response, indicating no existing targets match this pattern.

    {
      "indices": [ ],
      "aliases": [ ],
      "data_streams": [ ]
    }
  2. Create or update an index template. This template should contain the mappings and settings you’d like to apply to the new data stream’s backing indices.

    This index template must meet the requirements for a data stream template. It should also contain your previously chosen name or index pattern in the index_patterns property.

    If you are only adding or changing a few things, we recommend you create a new template by copying an existing one and modifying it as needed.

    For example, my-data-stream-template is an existing index template used by my-data-stream.

    The following create or update index template API request creates a new index template, new-data-stream-template. new-data-stream-template uses my-data-stream-template as its basis, with the following changes:

    • The index pattern in index_patterns matches any index or data stream starting with new-data-stream.
    • The @timestamp field mapping uses the date_nanos field data type rather than the date data type.
    • The template includes sort.field and sort.order index settings, which were not in the original my-data-stream-template template.
    response = client.indices.put_index_template(
      name: 'new-data-stream-template',
      body: {
        index_patterns: [
          'new-data-stream*'
        ],
        data_stream: {},
        priority: 500,
        template: {
          mappings: {
            properties: {
              "@timestamp": {
                type: 'date_nanos'
              }
            }
          },
          settings: {
            'sort.field' => [
              '@timestamp'
            ],
            'sort.order' => [
              'desc'
            ]
          }
        }
      }
    )
    puts response
    PUT /_index_template/new-data-stream-template
    {
      "index_patterns": [ "new-data-stream*" ],
      "data_stream": { },
      "priority": 500,
      "template": {
        "mappings": {
          "properties": {
            "@timestamp": {
              "type": "date_nanos"                 
            }
          }
        },
        "settings": {
          "sort.field": [ "@timestamp"],          
          "sort.order": [ "desc"]                 
        }
      }
    }

    Changes the @timestamp field mapping to the date_nanos field data type.

    Adds the sort.field index setting.

    Adds the sort.order index setting.

  3. Use the create data stream API to manually create the new data stream. The name of the data stream must match the index pattern defined in the new template’s index_patterns property.

    We do not recommend indexing new data to create this data stream. Later, you will reindex older data from an existing data stream into this new stream. This could result in one or more backing indices that contains a mix of new and old data.

    Mixing new and old data in a data stream

    While mixing new and old data is safe, it could interfere with data retention. If you delete older indices, you could accidentally delete a backing index that contains both new and old data. To prevent premature data loss, you would need to retain such a backing index until you are ready to delete its newest data.

    The following create data stream API request targets new-data-stream, which matches the index pattern for new-data-stream-template. Because no existing index or data stream uses this name, this request creates the new-data-stream data stream.

    response = client.indices.create_data_stream(
      name: 'new-data-stream'
    )
    puts response
    PUT /_data_stream/new-data-stream
  4. If you do not want to mix new and old data in your new data stream, pause the indexing of new documents. While mixing old and new data is safe, it could interfere with data retention. See Mixing new and old data in a data stream.
  5. If you use ILM to automate rollover, reduce the ILM poll interval. This ensures the current write index doesn’t grow too large while waiting for the rollover check. By default, ILM checks rollover conditions every 10 minutes.

    The following cluster update settings API request lowers the indices.lifecycle.poll_interval setting to 1m (one minute).

    response = client.cluster.put_settings(
      body: {
        persistent: {
          'indices.lifecycle.poll_interval' => '1m'
        }
      }
    )
    puts response
    PUT /_cluster/settings
    {
      "persistent": {
        "indices.lifecycle.poll_interval": "1m"
      }
    }
  6. Reindex your data to the new data stream using an op_type of create.

    If you want to partition the data in the order in which it was originally indexed, you can run separate reindex requests. These reindex requests can use individual backing indices as the source. You can use the get data stream API to retrieve a list of backing indices.

    For example, you plan to reindex data from my-data-stream into new-data-stream. However, you want to submit a separate reindex request for each backing index in my-data-stream, starting with the oldest backing index. This preserves the order in which the data was originally indexed.

    The following get data stream API request retrieves information about my-data-stream, including a list of its backing indices.

    response = client.indices.get_data_stream(
      name: 'my-data-stream'
    )
    puts response
    GET /_data_stream/my-data-stream

    The response’s indices property contains an array of the stream’s current backing indices. The first item in the array contains information about the stream’s oldest backing index.

    {
      "data_streams": [
        {
          "name": "my-data-stream",
          "timestamp_field": {
            "name": "@timestamp"
          },
          "indices": [
            {
              "index_name": ".ds-my-data-stream-2099.03.07-000001", 
              "index_uuid": "Gpdiyq8sRuK9WuthvAdFbw",
              "prefer_ilm": true,
              "managed_by": "Unmanaged"
            },
            {
              "index_name": ".ds-my-data-stream-2099.03.08-000002",
              "index_uuid": "_eEfRrFHS9OyhqWntkgHAQ",
              "prefer_ilm": true,
              "managed_by": "Unmanaged"
            }
          ],
          "generation": 2,
          "status": "GREEN",
          "next_generation_managed_by": "Unmanaged",
          "prefer_ilm": true,
          "template": "my-data-stream-template",
          "hidden": false,
          "system": false,
          "allow_custom_routing": false,
          "replicated": false,
          "rollover_on_write": false
        }
      ]
    }

    First item in the indices array for my-data-stream. This item contains information about the stream’s oldest backing index, .ds-my-data-stream-2099.03.07-000001.

    The following reindex API request copies documents from .ds-my-data-stream-2099.03.07-000001 to new-data-stream. The request’s op_type is create.

    response = client.reindex(
      body: {
        source: {
          index: '.ds-my-data-stream-2099.03.07-000001'
        },
        dest: {
          index: 'new-data-stream',
          op_type: 'create'
        }
      }
    )
    puts response
    POST /_reindex
    {
      "source": {
        "index": ".ds-my-data-stream-2099.03.07-000001"
      },
      "dest": {
        "index": "new-data-stream",
        "op_type": "create"
      }
    }

    You can also use a query to reindex only a subset of documents with each request.

    The following reindex API request copies documents from my-data-stream to new-data-stream. The request uses a range query to only reindex documents with a timestamp within the last week. Note the request’s op_type is create.

    response = client.reindex(
      body: {
        source: {
          index: 'my-data-stream',
          query: {
            range: {
              "@timestamp": {
                gte: 'now-7d/d',
                lte: 'now/d'
              }
            }
          }
        },
        dest: {
          index: 'new-data-stream',
          op_type: 'create'
        }
      }
    )
    puts response
    POST /_reindex
    {
      "source": {
        "index": "my-data-stream",
        "query": {
          "range": {
            "@timestamp": {
              "gte": "now-7d/d",
              "lte": "now/d"
            }
          }
        }
      },
      "dest": {
        "index": "new-data-stream",
        "op_type": "create"
      }
    }
  7. If you previously changed your ILM poll interval, change it back to its original value when reindexing is complete. This prevents unnecessary load on the master node.

    The following cluster update settings API request resets the indices.lifecycle.poll_interval setting to its default value.

    response = client.cluster.put_settings(
      body: {
        persistent: {
          'indices.lifecycle.poll_interval' => nil
        }
      }
    )
    puts response
    PUT /_cluster/settings
    {
      "persistent": {
        "indices.lifecycle.poll_interval": null
      }
    }
  8. Resume indexing using the new data stream. Searches on this stream will now query your new data and the reindexed data.
  9. Once you have verified that all reindexed data is available in the new data stream, you can safely remove the old stream.

    The following delete data stream API request deletes my-data-stream. This request also deletes the stream’s backing indices and any data they contain.

    response = client.indices.delete_data_stream(
      name: 'my-data-stream'
    )
    puts response
    DELETE /_data_stream/my-data-stream

Update or add an alias to a data stream

edit

Use the aliases API to update an existing data stream’s aliases. Changing an existing data stream’s aliases in its index pattern has no effect.

For example, the logs alias points to a single data stream. The following request swaps the stream for the alias. During this swap, the logs alias has no downtime and never points to both streams at the same time.

response = client.indices.update_aliases(
  body: {
    actions: [
      {
        remove: {
          index: 'logs-nginx.access-prod',
          alias: 'logs'
        }
      },
      {
        add: {
          index: 'logs-my_app-default',
          alias: 'logs'
        }
      }
    ]
  }
)
puts response
POST _aliases
{
  "actions": [
    {
      "remove": {
        "index": "logs-nginx.access-prod",
        "alias": "logs"
      }
    },
    {
      "add": {
        "index": "logs-my_app-default",
        "alias": "logs"
      }
    }
  ]
}