Configure the Elasticsearch output

The enabled config is a boolean setting to enable or disable the output. If set to false, the output is disabled.

The default value is true.

The list of Elasticsearch nodes to connect to. The events are distributed to these nodes in round robin order. If one node becomes unreachable, the event is automatically sent to another node. Each Elasticsearch node can be defined as a URL or IP:PORT. For example: http://192.15.3.2, https://es.found.io:9230 or 192.24.3.2:9300. If no port is specified, 9200 is used.

output.elasticsearch:
  hosts: ["10.45.3.2:9220", "10.45.3.1:9230"] 
  protocol: https
  path: /elasticsearch

In the previous example, the Elasticsearch nodes are available at https://10.45.3.2:9220/elasticsearch and https://10.45.3.1:9230/elasticsearch.

The gzip compression level. Setting this value to 0 disables compression. The compression level must be in the range of 1 (best speed) to 9 (best compression).

Increasing the compression level will reduce the network usage but will increase the cpu usage.

The default value is 1.

Configure escaping of HTML in strings. Set to true to enable escaping.

The default value is false.

The number of workers per configured host publishing events to Elasticsearch. This is best used with load balancing mode enabled. Example: If you have 2 hosts and 3 workers, in total 6 workers are started (3 for each host).

The default value is 1.

When loadbalance: true is set, Metricbeat connects to all configured hosts and sends data through all connections in parallel. If a connection fails, data is sent to the remaining hosts until it can be reestablished. Data will still be sent as long as Metricbeat can connect to at least one of its configured hosts.

When loadbalance: false is set, Metricbeat sends data to a single host at a time. The target host is chosen at random from the list of configured hosts, and all data is sent to that target until the connection fails, when a new target is selected. Data will still be sent as long as Metricbeat can connect to at least one of its configured hosts.

The default value is true.

output.elasticsearch:
  hosts: ["localhost:9200", "localhost:9201"]
  loadbalance: true

Instead of using a username and password, you can use API keys to secure communication with Elasticsearch. The value must be the ID of the API key and the API key joined by a colon: id:api_key.

See Grant access using API keys for more information.

The basic authentication username for connecting to Elasticsearch.

This user needs the privileges required to publish events to Elasticsearch. To create a user like this, see Create a publishing user.

The basic authentication password for connecting to Elasticsearch.

Dictionary of HTTP parameters to pass within the url with index operations.

The name of the protocol Elasticsearch is reachable on. The options are: http or https. The default is http. However, if you specify a URL for hosts, the value of protocol is overridden by whatever scheme you specify in the URL.

An HTTP path prefix that is prepended to the HTTP API calls. This is useful for the cases where Elasticsearch listens behind an HTTP reverse proxy that exports the API under a custom prefix.

Custom HTTP headers to add to each request created by the Elasticsearch output. Example:

output.elasticsearch.headers:
  X-My-Header: Header contents

It is possible to specify multiple header values for the same header name by separating them with a comma.

If set to true all proxy settings, including HTTP_PROXY and HTTPS_PROXY variables are ignored.

The URL of the proxy to use when connecting to the Elasticsearch servers. The value must be a complete URL. If a value is not specified through the configuration file then proxy environment variables are used. See the Go documentation for more information about the environment variables.

Additional headers to send to proxies during CONNECT requests.

The indexing target to write events to. Can point to an index, alias, or data stream. When using daily indices, this will be the index name. The default is "metricbeat-%{[agent.version]}-%{+yyyy.MM.dd}", for example, "metricbeat-8.16.1-2024-11-22". If you change this setting, you also need to configure the setup.template.name and setup.template.pattern options (see Elasticsearch index template).

If you are using the pre-built Kibana dashboards, you also need to set the setup.dashboards.index option (see Kibana dashboards).

When index lifecycle management (ILM) is enabled, the default index is "metricbeat-%{[agent.version]}-%{+yyyy.MM.dd}-%{index_num}", for example, "metricbeat-8.16.1-2024-11-22-000001". Custom index settings are ignored when ILM is enabled. If you’re sending events to a cluster that supports index lifecycle management, see Index lifecycle management (ILM) to learn how to change the index name.

You can set the index dynamically by using a format string to access any event field. For example, this configuration uses a custom field, fields.log_type, to set the index:

output.elasticsearch:
  hosts: ["http://localhost:9200"]
  index: "%{[fields.log_type]}-%{[agent.version]}-%{+yyyy.MM.dd}" 

With this configuration, all events with log_type: normal are sent to an index named normal-8.16.1-2024-11-22, and all events with log_type: critical are sent to an index named critical-8.16.1-2024-11-22.

See the indices setting for other ways to set the index dynamically.

An array of index selector rules. Each rule specifies the index to use for events that match the rule. During publishing, Metricbeat uses the first matching rule in the array. Rules can contain conditionals, format string-based fields, and name mappings. If the indices setting is missing or no rule matches, the index setting is used.

Similar to index, defining custom indices will disable Index lifecycle management (ILM).

Rule settings:

The following example sets the index based on whether the message field contains the specified string:

output.elasticsearch:
  hosts: ["http://localhost:9200"]
  indices:
    - index: "warning-%{[agent.version]}-%{+yyyy.MM.dd}"
      when.contains:
        message: "WARN"
    - index: "error-%{[agent.version]}-%{+yyyy.MM.dd}"
      when.contains:
        message: "ERR"

This configuration results in indices named warning-8.16.1-2024-11-22 and error-8.16.1-2024-11-22 (plus the default index if no matches are found).

The following example sets the index by taking the name returned by the index format string and mapping it to a new name that’s used for the index:

output.elasticsearch:
  hosts: ["http://localhost:9200"]
  indices:
    - index: "%{[fields.log_type]}"
      mappings:
        critical: "sev1"
        normal: "sev2"
      default: "sev3"

This configuration results in indices named sev1, sev2, and sev3.

The mappings setting simplifies the configuration, but is limited to string values. You cannot specify format strings within the mapping pairs.

Configuration options for index lifecycle management.

See Index lifecycle management (ILM) for more information.

A format string value that specifies the ingest pipeline to write events to.

output.elasticsearch:
  hosts: ["http://localhost:9200"]
  pipeline: my_pipeline_id

For more information, see Parse data using an ingest pipeline.

You can set the ingest pipeline dynamically by using a format string to access any event field. For example, this configuration uses a custom field, fields.log_type, to set the pipeline for each event:

output.elasticsearch:
  hosts: ["http://localhost:9200"]
  pipeline: "%{[fields.log_type]}_pipeline"

With this configuration, all events with log_type: normal are sent to a pipeline named normal_pipeline, and all events with log_type: critical are sent to a pipeline named critical_pipeline.

See the pipelines setting for other ways to set the ingest pipeline dynamically.

An array of pipeline selector rules. Each rule specifies the ingest pipeline to use for events that match the rule. During publishing, Metricbeat uses the first matching rule in the array. Rules can contain conditionals, format string-based fields, and name mappings. If the pipelines setting is missing or no rule matches, the pipeline setting is used.

Rule settings:

The following example sends events to a specific pipeline based on whether the message field contains the specified string:

output.elasticsearch:
  hosts: ["http://localhost:9200"]
  pipelines:
    - pipeline: "warning_pipeline"
      when.contains:
        message: "WARN"
    - pipeline: "error_pipeline"
      when.contains:
        message: "ERR"

The following example sets the pipeline by taking the name returned by the pipeline format string and mapping it to a new name that’s used for the pipeline:

output.elasticsearch:
  hosts: ["http://localhost:9200"]
  pipelines:
    - pipeline: "%{[fields.log_type]}"
      mappings:
        critical: "sev1_pipeline"
        normal: "sev2_pipeline"
      default: "sev3_pipeline"

With this configuration, all events with log_type: critical are sent to sev1_pipeline, all events with log_type: normal are sent to a sev2_pipeline, and all other events are sent to sev3_pipeline.

For more information about ingest pipelines, see Parse data using an ingest pipeline.

The number of times to retry publishing an event after a publishing failure. After the specified number of retries, the events are typically dropped.

Set max_retries to a value less than 0 to retry until all events are published.

The default is 3.

The maximum number of events to bulk in a single Elasticsearch bulk API index request. The default is 1600.

Events can be collected into batches. Metricbeat will split batches read from the queue which are larger than bulk_max_size into multiple batches.

Specifying a larger batch size can improve performance by lowering the overhead of sending events. However big batch sizes can also increase processing times, which might result in API errors, killed connections, timed-out publishing requests, and, ultimately, lower throughput.

Setting bulk_max_size to values less than or equal to 0 disables the splitting of batches. When splitting is disabled, the queue decides on the number of events to be contained in a batch.

The number of seconds to wait before trying to reconnect to Elasticsearch after a network error. After waiting backoff.init seconds, Metricbeat tries to reconnect. If the attempt fails, the backoff timer is increased exponentially up to backoff.max. After a successful connection, the backoff timer is reset. The default is 1s.

The maximum number of seconds to wait before attempting to connect to Elasticsearch after a network error. The default is 60s.

The maximum amount of time an idle connection will remain idle before closing itself. Zero means no limit. The format is a Go language duration (example 60s is 60 seconds). The default is 3s.

The http request timeout in seconds for the Elasticsearch request. The default is 90.

By default, Metricbeat expects the Elasticsearch instance to be on the same or newer version to provide optimal experience. We suggest you connect to the same version to make sure all features Metricbeat is using are available in your Elasticsearch instance.

You can disable the check for example during updating the Elastic Stack, so data collection can go on.

Configuration options for SSL parameters like the certificate authority to use for HTTPS-based connections. If the ssl section is missing, the host CAs are used for HTTPS connections to Elasticsearch.

See the secure communication with Elasticsearch guide or SSL configuration reference for more information.

Configuration options for Kerberos authentication.

See Kerberos for more information.

Specifies the behavior when the elasticsearch cluster explicitly rejects documents, for example on mapping conflicts.

output.elasticsearch:
  hosts: ["http://localhost:9200"]
  non_indexable_policy.drop: ~

output.elasticsearch:
  hosts: ["http://localhost:9200"]
  non_indexable_policy.dead_letter_index:
    index: "my-dead-letter-index"

The performance preset to apply to the output configuration.

output.elasticsearch:
  hosts: ["http://localhost:9200"]
  preset: balanced

Performance presets apply a set of configuration overrides based on a desired performance goal. If set, a performance preset will override other configuration flags to match the recommended settings for that preset. Valid options are: * balanced: good starting point for general efficiency * throughput: good for high data volumes, may increase cpu and memory requirements * scale: reduces ambient resource use in large low-throughput deployments * latency: minimize the time for fresh data to become visible in Elasticsearch * custom: apply user configuration directly with no overrides

The default if unspecified is custom.

Presets represent current recommendations based on the intended goal; their effect may change between versions to better suit those goals. Currently the presets have the following effects: