Menu

Caution

Grafana Alloy is the new name for our distribution of the OTel collector. Grafana Agent has been deprecated and is in Long-Term Support (LTS) through October 31, 2025. Grafana Agent will reach an End-of-Life (EOL) on November 1, 2025. Read more about why we recommend migrating to Grafana Alloy.

Important: This documentation is about an older version. It's relevant only to the release noted, many of the features and functions have been updated or replaced. Please view the current version.

Open source

otelcol.processor.batch

otelcol.processor.batch accepts telemetry data from other otelcol components and places them into batches. Batching improves the compression of data and reduces the number of outgoing network requests required to transmit data. This processor supports both size and time based batching.

We strongly recommend that you configure the batch processor on every Agent that uses OpenTelemetry (otelcol) Flow components. The batch processor should be defined in the pipeline after the otelcol.processor.memory_limiter as well as any sampling processors. This is because batching should happen after any data drops such as sampling.

NOTE: otelcol.processor.batch is a wrapper over the upstream OpenTelemetry Collector batch processor. Bug reports or feature requests will be redirected to the upstream repository, if necessary.

Multiple otelcol.processor.batch components can be specified by giving them different labels.

Usage

river
otelcol.processor.batch "LABEL" {
  output {
    metrics = [...]
    logs    = [...]
    traces  = [...]
  }
}

Arguments

otelcol.processor.batch supports the following arguments:

NameTypeDescriptionDefaultRequired
timeoutdurationHow long to wait before flushing the batch."200ms"no
send_batch_sizenumberAmount of data to buffer before flushing the batch.8192no
send_batch_max_sizenumberUpper limit of a batch size.0no
metadata_keyslist(string)Creates a different batcher for each key/value combination of metadata.[]no
metadata_cardinality_limitnumberLimit of the unique metadata key/value combinations.1000no

otelcol.processor.batch accumulates data into a batch until one of the following events happens:

  • The duration specified by timeout elapses since the time the last batch was sent.

  • The number of spans, log lines, or metric samples processed is greater than or equal to the number specified by send_batch_size.

Logs, traces, and metrics are processed independently. For example, if send_batch_size is set to 1000:

  • The processor may, at the same time, buffer 1,000 spans, 1,000 log lines, and 1,000 metric samples before flushing them.
  • If there are enough spans for a batch of spans (1,000 or more), but not enough for a batch of metric samples (less than 1,000) then only the spans will be flushed.

Use send_batch_max_size to limit the amount of data contained in a single batch:

  • When set to 0, batches can be any size.
  • When set to a non-zero value, send_batch_max_size must be greater than or equal to send_batch_size. Every batch will contain up to the send_batch_max_size number of spans, log lines, or metric samples. The excess spans, log lines, or metric samples will not be lost - instead, they will be added to the next batch.

For example, assume send_batch_size is set to the default 8192 and there are currently 8,000 batched spans. If the batch processor receives 8,000 more spans at once, its behavior depends on how send_batch_max_size is configured:

  • If send_batch_max_size is set to 0, the total batch size would be 16,000 which would then be flushed as a single batch.
  • If send_batch_max_size is set to 10000, then the total batch size will be 10,000 and the remaining 6,000 spans will be flushed in a subsequent batch.

metadata_cardinality_limit applies for the lifetime of the process.

Receivers should be configured with include_metadata = true so that metadata keys are available to the processor.

Each distinct combination of metadata triggers the allocation of a new background task in the Agent that runs for the lifetime of the process, and each background task holds one pending batch of up to send_batch_size records. Batching by metadata can therefore substantially increase the amount of memory dedicated to batching.

The maximum number of distinct combinations is limited to the configured metadata_cardinality_limit, which defaults to 1000 to limit memory impact.

Blocks

The following blocks are supported inside the definition of otelcol.processor.batch:

HierarchyBlockDescriptionRequired
outputoutputConfigures where to send received telemetry data.yes

output block

The output block configures a set of components to forward resulting telemetry data to.

The following arguments are supported:

NameTypeDescriptionDefaultRequired
logslist(otelcol.Consumer)List of consumers to send logs to.[]no
metricslist(otelcol.Consumer)List of consumers to send metrics to.[]no
traceslist(otelcol.Consumer)List of consumers to send traces to.[]no

You must specify the output block, but all its arguments are optional. By default, telemetry data is dropped. Configure the metrics, logs, and traces arguments accordingly to send telemetry data to other components.

Exported fields

The following fields are exported and can be referenced by other components:

NameTypeDescription
inputotelcol.ConsumerA value that other components can use to send telemetry data to.

input accepts otelcol.Consumer data for any telemetry signal (metrics, logs, or traces).

Component health

otelcol.processor.batch is only reported as unhealthy if given an invalid configuration.

Debug information

otelcol.processor.batch does not expose any component-specific debug information.

Debug metrics

  • processor_batch_batch_send_size_ratio (histogram): Number of units in the batch.
  • processor_batch_metadata_cardinality_ratio (gauge): Number of distinct metadata value combinations being processed.
  • processor_batch_timeout_trigger_send_ratio_total (counter): Number of times the batch was sent due to a timeout trigger.
  • processor_batch_batch_size_trigger_send_ratio_total (counter): Number of times the batch was sent due to a size trigger.

Examples

Basic usage

This example batches telemetry data before sending it to otelcol.exporter.otlp for further processing:

river
otelcol.processor.batch "default" {
  output {
    metrics = [otelcol.exporter.otlp.production.input]
    logs    = [otelcol.exporter.otlp.production.input]
    traces  = [otelcol.exporter.otlp.production.input]
  }
}

otelcol.exporter.otlp "production" {
  client {
    endpoint = env("OTLP_SERVER_ENDPOINT")
  }
}

Batching with a timeout

This example will buffer up to 10,000 spans, metric data points, or log records for up to 10 seconds. Because send_batch_max_size is not set, the batch size may exceed 10,000.

river
otelcol.processor.batch "default" {
  timeout = "10s"
  send_batch_size = 10000

  output {
    metrics = [otelcol.exporter.otlp.production.input]
    logs    = [otelcol.exporter.otlp.production.input]
    traces  = [otelcol.exporter.otlp.production.input]
  }
}

otelcol.exporter.otlp "production" {
  client {
    endpoint = env("OTLP_SERVER_ENDPOINT")
  }
}

Batching based on metadata

Batching by metadata enables support for multi-tenant OpenTelemetry pipelines with batching over groups of data having the same authorization metadata.

river
otelcol.receiver.jaeger "default" {
  protocols {
    grpc {
      include_metadata = true
    }
    thrift_http {}
    thrift_binary {}
    thrift_compact {}
  }

  output {
    traces = [otelcol.processor.batch.default.input]
  }
}

otelcol.processor.batch "default" {
  // batch data by tenant id
  metadata_keys = ["tenant_id"]
  // limit to 10 batcher processes before raising errors
  metadata_cardinality_limit = 123

  output {
    traces  = [otelcol.exporter.otlp.production.input]
  }
}

otelcol.exporter.otlp "production" {
  client {
    endpoint = env("OTLP_SERVER_ENDPOINT")
  }
}

Compatible components

otelcol.processor.batch can accept arguments from the following components:

otelcol.processor.batch has exports that can be consumed by the following components:

Note

Connecting some components may not be sensible or components may require further configuration to make the connection work correctly. Refer to the linked documentation for more details.