Inferred services

Docs > APM > Service Observability > Inferred services

Overview

Datadog automatically discovers the dependencies for an instrumented service, such as databases, queues, or third-party APIs, even if that dependency hasn’t been directly instrumented. By analyzing outbound requests from your instrumented services, Datadog infers the presence of these dependencies and collects associated performance metrics.

Explore inferred services in the Service Catalog by filtering entries by entity type, such as database, queue, or third-party API. Each service page is tailored to the type of service you are investigating. For instance, database service pages show database-specific insights and include database monitoring data if you are using Database Monitoring.

Set up inferred services

To see inferred services, you must enable some configurations.

For Datadog Agent versions 7.55.1 or later, add the following to your datadog.yaml configuration file:

datadog.yaml

apm_config:
  compute_stats_by_span_kind: true
  peer_tags_aggregation: true

Alternatively, set these environment variables in your Datadog Agent configuration:

DD_APM_COMPUTE_STATS_BY_SPAN_KIND=true 
DD_APM_PEER_TAGS_AGGREGATION=true

If you are using Helm, include these environment variables in your values.yaml file.

For Datadog Agent versions 7.50.3 through 7.54.1, add the following to your datadog.yaml configuration file:

datadog.yaml

apm_config:
  compute_stats_by_span_kind: true
  peer_tags_aggregation: true
  peer_tags: ["_dd.base_service","amqp.destination","amqp.exchange","amqp.queue","aws.queue.name","aws.s3.bucket","bucketname","cassandra.keyspace","db.cassandra.contact.points","db.couchbase.seed.nodes","db.hostname","db.instance","db.name","db.namespace","db.system","grpc.host","hostname","http.host","http.server_name","messaging.destination","messaging.destination.name","messaging.kafka.bootstrap.servers","messaging.rabbitmq.exchange","messaging.system","mongodb.db","msmq.queue.path","net.peer.name","network.destination.name","peer.hostname","peer.service","queuename","rpc.service","rpc.system","server.address","streamname","tablename","topicname"]

Alternatively, set these environment variables in your Datadog Agent configuration:

DD_APM_COMPUTE_STATS_BY_SPAN_KIND=true 
DD_APM_PEER_TAGS_AGGREGATION=true
DD_APM_PEER_TAGS='["_dd.base_service","amqp.destination","amqp.exchange","amqp.queue","aws.queue.name","aws.s3.bucket","bucketname","cassandra.keyspace","db.cassandra.contact.points","db.couchbase.seed.nodes","db.hostname","db.instance","db.name","db.namespace","db.system","grpc.host","hostname","http.host","http.server_name","messaging.destination","messaging.destination.name","messaging.kafka.bootstrap.servers","messaging.rabbitmq.exchange","messaging.system","mongodb.db","msmq.queue.path","net.peer.name","network.destination.name","peer.hostname","peer.service","queuename","rpc.service","rpc.system","server.address","streamname","tablename","topicname"]'

If you are using Helm, include these environment variables in your values.yaml file.

For the OpenTelemetry Collector, the minimum recommended version is opentelemetry-collector-contrib v0.95.0 or later. In that case, update this configuration:

connectors:
  datadog/connector:
    traces:
      compute_stats_by_span_kind: true
      peer_tags_aggregation: true
      peer_tags: ["_dd.base_service","amqp.destination","amqp.exchange","amqp.queue","aws.queue.name","aws.s3.bucket","bucketname","db.cassandra.contact.points","db.couchbase.seed.nodes","db.hostname","db.instance","db.name","db.namespace","db.system","grpc.host","hostname","http.host","http.server_name","messaging.destination","messaging.destination.name","messaging.kafka.bootstrap.servers","messaging.rabbitmq.exchange","messaging.system","mongodb.db","msmq.queue.path","net.peer.name","network.destination.name","peer.hostname","peer.service","queuename","rpc.service","rpc.system","server.address","streamname","tablename","topicname"]

If your Collector version is below v0.95.0, update the following Collector configuration:

exporters:
  datadog:
    traces:
      compute_stats_by_span_kind: true
      peer_tags_aggregation: true
      peer_tags: ["_dd.base_service","amqp.destination","amqp.exchange","amqp.queue","aws.queue.name","aws.s3.bucket","bucketname","db.cassandra.contact.points","db.couchbase.seed.nodes","db.hostname","db.instance","db.name","db.namespace","db.system","grpc.host","hostname","http.host","http.server_name","messaging.destination","messaging.destination.name","messaging.kafka.bootstrap.servers","messaging.rabbitmq.exchange","messaging.system","mongodb.db","msmq.queue.path","net.peer.name","network.destination.name","peer.hostname","peer.service","queuename","rpc.service","rpc.system","server.address","streamname","tablename","topicname"]   

Example: collector.yaml.

Naming inferred entities

To determine the names and types of the inferred service dependencies, Datadog uses standard span attributes and maps them to peer.* attributes. For example, inferred external APIs use the default naming scheme net.peer.name like api.stripe.com, api.twilio.com, and us6.api.mailchimp.com. Inferred databases use the default naming scheme db.instance.

Peer tags

Peer Tag	Source Attributes
`peer.aws.dynamodb.table`	`tablename`
`peer.aws.kinesis.stream`	`streamname`
`peer.aws.s3.bucket`	`bucketname`, `aws.s3.bucket`
`peer.aws.sqs.queue`	`queuename`
`peer.cassandra.contact.points`	`db.cassandra.contact.points`
`peer.couchbase.seed.nodes`	`db.couchbase.seed.nodes`
`peer.db.name`	`db.name`, `mongodb.db`, `db.instance`, `cassandra.keyspace`, `db.namespace`
`peer.db.system`	`db.system`
`peer.hostname`	`peer.hostname`, `hostname`, `net.peer.name`, `db.hostname`, `network.destination.name`, `grpc.host`, `http.host`, `server.address`, `http.server_name`
`peer.kafka.bootstrap.servers`	`messaging.kafka.bootstrap.servers`
`peer.messaging.destination`	`topicname`, `messaging.destination`, `messaging.destination.name`, `messaging.rabbitmq.exchange`, `amqp.destination`, `amqp.queue`, `amqp.exchange`, `msmq.queue.path`, `aws.queue.name`
`peer.messaging.system`	`messaging.system`
`peer.rpc.service`	`rpc.service`
`peer.rpc.system`	`rpc.system`
`peer.service`	`peer.service`

Note: Peer attribute values that match IP address formats (for example, 127.0.0.1) are modified and redacted with blocked-ip-address to prevent unnecessary noise and tagging metrics with high-cardinality dimensions. As a result, you may encounter some blocked-ip-address services appearing as downstream dependencies of your instrumented services.

Migrate to global default service naming

With inferred services, service dependencies are automatically detected from existing span attributes. As a result, changing service names (using the service tag) is not required to identify these dependencies.

Enable DD_TRACE_REMOVE_INTEGRATION_SERVICE_NAMES_ENABLED to ensures no Datadog integration sets service names that are different from the default global service name. This also improves how service-to-service connections and inferred services are represented in Datadog visualizations, across all supported tracing library languages and integrations.

Enabling this option may impact existing APM metrics, custom span metrics, trace analytics, retention filters, sensitive data scans, monitors, dashboards, or notebooks that reference the old service names. Update these assets to use the global default service tag (service:<DD_SERVICE>).

For instructions on how to remove service overrides and migrate to inferred services, see the Service Overrides guide.