Filter Streamed Telemetry

By default, a ThousandEyes for OpenTelemetry integration exports all supported telemetry for its selected signal, data model version, and data sources. You can control the exported data by excluding specific metrics or attributes from an integration.

Common reasons to exclude attributes include:

• Reducing telemetry cardinality.

• Preventing specific supported attributes that contain personally identifiable information (PII) from being exported.

Filters match attribute names, not individual attribute values.

Filtering is configured separately for each integration. Metric and attribute names are matched exactly and are case-sensitive.

Filter Behavior

The available filters depend on the integration signal:

Signal	Metric exclusions	Attribute exclusions
Metrics	Supported for v1 and v2	Supported for v1 and v2
Traces	Not supported	Supported for v2
Logs	Not supported	Supported for v2

Metric exclusions remove each matching metric and all of its data points from the integration output.

Attribute exclusions remove each matching top-level attribute wherever that attribute is supported for the signal:

• Metrics: Resource, instrumentation scope, and data point attributes.

• Traces: Resource, instrumentation scope, and span attributes. Span event and span link attributes are not filtered.

• Logs: Resource, instrumentation scope, and log record attributes.

Excluding an attribute does not remove its resource, span, log record, metric, or data point. If an excluded attribute is not present in a telemetry item, that item is unchanged.

Filtering changes may take a short time to appear in exported telemetry while the integration configuration is refreshed.

Supported Metric and Attribute Names

The filter options are based on the integration's signal and data model version. Use the metric and attribute names documented in the corresponding data model:

Data model and signal	Catalog
Data model v1 metrics	Metrics and attributes
Data model v2 metrics	Metrics and attributes
Data model v2 traces	Trace attributes
Data model v2 logs	Log attributes

The following attributes are reserved and cannot be excluded:

Data model version	Reserved attributes
v1	account.id, thousandeyes.data.version, thousandeyes.stream.id
v2	thousandeyes.account.id, thousandeyes.data.version, thousandeyes.stream.id

The UI does not display reserved attributes in its selectors. API requests can contain only names supported for the selected signal and data model version. Names that are not in the supported filter catalog, including attributes created dynamically from tags, are rejected.

Configure Filtering Using the UI

You can configure filtering while creating an integration or operation, or by editing an existing one.

Integrations 1.0

1. In the ThousandEyes platform, go to Manage > Integrations > Integrations 1.0.

2. Create a ThousandEyes for OpenTelemetry integration, or select an existing integration and click Edit.

3. If you are creating an integration, select the OpenTelemetry Signal and Data Model Version. If you are editing an existing integration, verify its read-only OpenTelemetry Signal and Data Model Version values. These values determine which filter options are available.

4. In Filter Configuration, select the data to exclude:

   • For a metric stream, use Metrics to exclude metrics and Specific Metric Attributes to exclude attributes.

   • For a trace or log stream, use Specific Attributes to exclude attributes.

5. Click Save.

Integrations 2.0

1. In the ThousandEyes platform, go to Manage > Integrations > Integrations 2.0.

2. Create or edit an OpenTelemetry operation and assign a connector.

3. If you are creating an operation, select the OpenTelemetry Signal. If you are editing an existing operation, verify its read-only OpenTelemetry Signal value. Integrations 2.0 filter options use data model v2.

4. In Filter Configuration, select the data to exclude:

   • For a metric stream, use Metrics to exclude metrics and Specific Metric Attributes to exclude attributes.

   • For a trace or log stream, use Specific Attributes to exclude attributes.

5. Click Save.

The selectors show how many supported names are excluded. If you leave a selector empty, the integration exports all supported metrics or attributes for that selector.

Configure Filtering Using the API

Use filters.metricsFilter.excludedMetrics and filters.attributesFilter.excludedAttributes when you create or update a stream. The following example creates a data model v2 metric stream that excludes network.loss and the server.address attribute:

curl -i -X POST https://api.thousandeyes.com/v7/streams \
  -H "Authorization: Bearer $BEARER_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "type": "opentelemetry",
    "name": "Filtered metrics stream",
    "signal": "metric",
    "endpointType": "http",
    "streamEndpointUrl": "https://example.com/v1/metrics",
    "dataModelVersion": "v2",
    "testMatch": [
      {
        "id": "987654",
        "domain": "cea"
      }
    ],
    "filters": {
      "metricsFilter": {
        "excludedMetrics": [
          "network.loss"
        ]
      },
      "attributesFilter": {
        "excludedAttributes": [
          "server.address"
        ]
      }
    }
  }'

For trace and log streams, omit metricsFilter and configure only attributesFilter.

When updating an integration with PUT /v7/streams/{id}:

• Omitting a filter preserves its current exclusions.

• Sending an empty excludedMetrics or excludedAttributes array clears the corresponding exclusions.

• Sending null as the value of excludedMetrics or excludedAttributes, an unsupported name, or a metric exclusion for a trace or log stream causes validation to fail.

For the complete endpoint schema and response definitions, see the ThousandEyes for OpenTelemetry API.