Semantic Conventions for Messaging Spans

Status: Experimental

Warning Existing Messaging instrumentations that are using v1.20.0 of this document (or prior):

• SHOULD NOT change the version of the networking conventions that they emit until the HTTP semantic conventions are marked stable (HTTP stabilization will include stabilization of a core set of networking conventions which are also used in Messaging instrumentations). Conventions include, but are not limited to, attributes, metric and span names, and unit of measure.
• SHOULD introduce an environment variable OTEL_SEMCONV_STABILITY_OPT_IN in the existing major version which is a comma-separated list of values. The only values defined so far are:
  • http - emit the new, stable networking conventions, and stop emitting the old experimental networking conventions that the instrumentation emitted previously.
  • http/dup - emit both the old and the stable networking conventions, allowing for a seamless transition.
  • The default behavior (in the absence of one of these values) is to continue emitting whatever version of the old experimental networking conventions the instrumentation was emitting previously.
  • Note: http/dup has higher precedence than http in case both values are present
• SHOULD maintain (security patching at a minimum) the existing major version for at least six months after it starts emitting both sets of conventions.
• SHOULD drop the environment variable in the next major version.

Definitions

Message

Although messaging systems are not as standardized as, e.g., HTTP, it is assumed that the following definitions are applicable to most of them that have similar concepts at all (names borrowed mostly from JMS):

A message is an envelope with a potentially empty body. This envelope may offer the possibility to convey additional metadata, often in key/value form.

A message is sent by a message producer to:

• Physically: some message broker (which can be e.g., a single server, or a cluster, or a local process reached via IPC). The broker handles the actual delivery, re-delivery, persistence, etc. In some messaging systems the broker may be identical or co-located with (some) message consumers. With Apache Kafka, the physical broker a message is written to depends on the number of partitions, and which broker is the leader of the partition the record is written to.
• Logically: some particular message destination.

Messages can be delivered to 0, 1, or multiple consumers depending on the dispatching semantic of the protocol.

Producer

The “producer” is a specific instance, process or device that creates and publishes a message. “Publishing” is the process of sending a message or batch of messages to the intermediary or consumer.

Consumer

A “consumer” receives the message and acts upon it. It uses the context and data to execute some logic, which might lead to the occurrence of new events.

The consumer receives, processes, and settles a message. “Receiving” is the process of obtaining a message from the intermediary, “processing” is the process of acting on the information a message contains, “settling” is the process of notifying an intermediary that a message was processed successfully.

Intermediary

An “intermediary” receives a message to forward it to the next receiver, which might be another intermediary or a consumer.

Destinations

A destination represents the entity within a messaging system where messages are published to and consumed from.

A destination is usually uniquely identified by its name within the messaging system instance. Examples of a destination name would be an URL or a simple one-word identifier.

In some use cases, messages are routed within one or multiple brokers. In such cases, the destination the message was originally published to is different from the destination it is being consumed from. When information about the destination where the message was originally published to is available, consumers can record them under the destination_publish namespace.

Typical examples of destinations include Kafka topics, RabbitMQ queues and topics.

Message consumption

The consumption of a message can happen in multiple steps. First, the lower-level receiving of a message at a consumer, and then the logical processing of the message. Often, the waiting for a message is not particularly interesting and hidden away in a framework that only invokes some handler function to process a message once one is received (in the same way that the listening on a TCP port for an incoming HTTP message is not particularly interesting).

Conversations

In some messaging systems, a message can receive one or more reply messages that answers a particular other message that was sent earlier. All messages that are grouped together by such a reply-relationship are called a conversation. The grouping usually happens through some sort of “In-Reply-To:” meta information or an explicit conversation ID (sometimes called correlation ID). Sometimes a conversation can span multiple message destinations (e.g. initiated via a topic, continued on a temporary one-to-one queue).

Temporary and anonymous destinations

Some messaging systems support the concept of temporary destination (often only temporary queues) that are established just for a particular set of communication partners (often one to one) or conversation. Often such destinations are also unnamed (anonymous) or have an auto-generated name.

Conventions

Given these definitions, the remainder of this section describes the semantic conventions for Spans describing interactions with messaging systems.

Context propagation

A message may traverse many different components and layers in one or more intermediaries when it is propagated from the producer to the consumer(s). To be able to correlate consumer traces with producer traces using the existing context propagation mechanisms, all components must propagate context down the chain.

Messaging systems themselves may trace messages as the messages travels from producers to consumers. Such tracing would cover the transport layer but would not help in correlating producers with consumers. To be able to directly correlate producers with consumers, another context that is propagated with the message is required.

A message creation context allows correlating producers with consumers of a message and model the dependencies between them, regardless of the underlying messaging transport mechanism and its instrumentation.

The message creation context is created by the producer and should be propagated to the consumer(s). Consumer traces cannot be directly correlated with producer traces if the message creation context is not attached and propagated with the message.

A producer SHOULD attach a message creation context to each message. If possible, the message creation context SHOULD be attached in such a way that it cannot be changed by intermediaries.

This document does not specify the exact mechanisms on how the creation context is attached/extracted to/from messages. Future versions of these conventions will give clear recommendations, following industry standards including, but not limited to Trace Context: AMQP protocol and Trace Context: MQTT protocol once those standards reach a stable state.

Span name

The span name SHOULD be set to the message destination name and the operation being performed in the following format:

<destination name> <operation name>

The destination name SHOULD only be used for the span name if it is known to be of low cardinality (cf. general span name guidelines). This can be assumed if it is statically derived from application code or configuration. Wherever possible, the real destination names after resolving logical or aliased names SHOULD be used. If the destination name is dynamic, such as a conversation ID or a value obtained from a Reply-To header, it SHOULD NOT be used for the span name. In these cases, an artificial destination name that best expresses the destination, or a generic, static fallback like "(anonymous)" for anonymous destinations SHOULD be used instead.

The values allowed for <operation name> are defined in the section Operation names below. If the format above is used, the operation name MUST match the messaging.operation attribute defined for message consumer spans below.

Examples:

• shop.orders publish
• shop.orders receive
• shop.orders process
• print_jobs publish
• topic with spaces process
• AuthenticationRequest-Conversations process
• (anonymous) publish ((anonymous) being a stable identifier for an unnamed destination)

Operation names

The following operations related to messages are defined for these semantic conventions:

Span kind

Span kinds SHOULD be set according to the following table, based on the operation a span describes.

For cases not covered by the table above, the span kind should be set according to the generic specification about span kinds, e. g. it should be set to CLIENT for the “Publish” span if its context is not used as creation context and if the “Publish” span models a synchronous call to the intermediary.

Setting span kinds according to this table ensures that span links between consumers and producers always exist between a PRODUCER span on the producer side and a CONSUMER span on the consumer side. This allows analysis tools to interpret linked traces without the need for additional semantic hints.

Trace structure

Producer spans

“Publish” spans SHOULD be created for operations of providing messages for sending or publishing to an intermediary. A single “Publish” span can account for a single message, or for multiple messages (in the case of providing messages in batches). “Create” spans MAY be created. A single “Create” span SHOULD account only for a single message. “Create” spans SHOULD either be children or links of the related “Publish” span.

If a user provides a custom creation context in a message, this context SHOULD NOT be modified, a “Create” span SHOULD NOT be created, and the “Publish” span SHOULD link to the custom creation context. Otherwise, if a “Create” span exists for a message, its context SHOULD be injected into the message. If no “Create” span exists and no custom creation context is injected into the message, the context of the related “Publish” span SHOULD be injected into the message.

Consumer spans

“Deliver” spans SHOULD be created for operations of passing messages to the application when those operations are not initiated by the application code (push-based scenarios). A “Deliver” span covers the duration of such an operation, which is usually a callback or handler.

“Receive” spans SHOULD be created for operations of passing messages to the application when those operations are initiated by the application code (pull-based scenarios).

“Deliver” or “Receive” spans MUST NOT be created for messages that are pre-fetched or cached by messaging libraries or SDKs until they are forwarded to the caller.

A single “Deliver” or “Receive” span can account for a single message, for a batch of messages, or for no message at all (if it is signalled that no messages were received). For each message it accounts for, the “Deliver” or “Receive” span SHOULD link to the message’s creation context.

Messaging attributes

[1]: The error.type SHOULD be predictable and SHOULD have low cardinality. Instrumentations SHOULD document the list of errors they report.

The cardinality of error.type within one instrumentation library SHOULD be low. Telemetry consumers that aggregate data from multiple instrumentation libraries and applications should be prepared for error.type to have high cardinality at query time when no additional filters are applied.

If the operation has completed successfully, instrumentations SHOULD NOT set error.type.

If a specific domain defines its own set of error identifiers (such as HTTP or gRPC status codes), it’s RECOMMENDED to:

• Use a domain-specific attribute
• Set error.type to capture all errors, regardless of whether they are defined within the domain-specific set or not.

[2]: If and only if the messaging operation has failed.

[3]: Instrumentations SHOULD NOT set messaging.batch.message_count on spans that operate with a single message. When a messaging client library supports both batch and single-message API for the same operation, instrumentations SHOULD use messaging.batch.message_count for batching APIs and SHOULD NOT use it for single-message APIs.

[4]: If the span describes an operation on a batch of messages.

[5]: If value is true. When missing, the value is assumed to be false.

[6]: Destination name SHOULD uniquely identify a specific queue, topic or other entity within the broker. If the broker doesn’t have such notion, the destination name SHOULD uniquely identify the broker.

[7]: If span describes operation on a single message or if the value applies to all messages in the batch.

[8]: Destination names could be constructed from templates. An example would be a destination name involving a user name or product id. Although the destination name in this case is of high cardinality, the underlying template is of low cardinality and can be effectively used for grouping and aggregation.

[9]: If available. Instrumentations MUST NOT use messaging.destination.name as template unless low-cardinality of destination name is guaranteed.

[10]: If value is true. When missing, the value is assumed to be false.

[11]: This can refer to both the compressed or uncompressed body size. If both sizes are known, the uncompressed body size should be used.

[12]: Only if span represents operation on a single message.

[13]: Only if span represents operation on a single message.

[14]: This can refer to both the compressed or uncompressed size. If both sizes are known, the uncompressed size should be used.

[15]: Only if span represents operation on a single message.

[16]: Only for spans that represent an operation on a single message.

[17]: If a custom value is used, it MUST be of low cardinality.

[18]: The value SHOULD be normalized to lowercase.

[19]: network.protocol.version refers to the version of the protocol used and might be different from the protocol client’s version. If the HTTP client has a version of 0.27.2, but sends HTTP version 1.1, this attribute should be set to 1.1.

[20]: The value SHOULD be normalized to lowercase.

Consider always setting the transport when setting a port number, since a port number is ambiguous without knowing the transport. For example different processes could be listening on TCP port 12345 and UDP port 12345.

[21]: The value SHOULD be normalized to lowercase.

[22]: This should be the IP/hostname of the broker (or other network-level peer) this specific message is sent to/received from.

[23]: When observed from the client side, and when communicating through an intermediary, server.port SHOULD represent the server port behind any intermediaries, for example proxies, if it’s available.

error.type has the following list of well-known values. If one of them applies, then the respective value MUST be used, otherwise a custom value MAY be used.

messaging.operation has the following list of well-known values. If one of them applies, then the respective value MUST be used, otherwise a custom value MAY be used.

messaging.system has the following list of well-known values. If one of them applies, then the respective value MUST be used, otherwise a custom value MAY be used.

network.transport has the following list of well-known values. If one of them applies, then the respective value MUST be used, otherwise a custom value MAY be used.

network.type has the following list of well-known values. If one of them applies, then the respective value MUST be used, otherwise a custom value MAY be used.

Additionally server.port from the network attributes is recommended. Furthermore, it is strongly recommended to add the network.transport attribute and follow its guidelines, especially for in-process queueing systems (like Hangfire, for example). These attributes should be set to the broker to which the message is sent/from which it is received.

Attribute namespaces

• messaging.message: Contains attributes that describe individual messages
• messaging.destination: Contains attributes that describe the logical entity messages are published to. See Destinations for more details
• messaging.destination_publish: Contains attributes that describe the logical entity messages were originally published to. See Destinations for more details
• messaging.batch: Contains attributes that describe batch operations
• messaging.consumer: Contains attributes that describe application instance that consumes a message. See consumer for more details

Communication with broker is described with general network attributes.

Messaging system-specific attributes MUST be defined in the corresponding messaging.{system} namespace as described in Attributes specific to certain messaging systems.

Consumer attributes

The following additional attributes describe message consumer operations.

Since messages could be routed by brokers, the destination messages are published to may not match with the destination they are consumed from.

If information about the original destination is available on the consumer, consumer instrumentations SHOULD populate the attributes under the namespace messaging.destination_publish.*

[1]: The name SHOULD uniquely identify a specific queue, topic, or other entity within the broker. If the broker doesn’t have such notion, the original destination name SHOULD uniquely identify the broker.

Per-message attributes

All messaging operations (publish, receive, process, or others not covered by this specification) can describe both single and/or batch of messages. Attributes in the messaging.message or messaging.{system}.message namespace describe individual messages. For single-message operations they SHOULD be set on corresponding span.

For batch operations, per-message attributes are usually different and cannot be set on the corresponding span. In such cases the attributes SHOULD be set on links. See Batch receiving for more information on correlation using links.

Some messaging systems (e.g., Kafka, Azure EventGrid) allow publishing a single batch of messages to different topics. In such cases, the attributes in messaging.destination MAY be set on links. Instrumentations MAY set destination attributes on the span if all messages in the batch share the same destination.

Attributes specific to certain messaging systems

All attributes that are specific for a messaging system SHOULD be populated in messaging.{system} namespace. Attributes that describe a message, a destination, a consumer, or a batch of messages SHOULD be populated under the corresponding namespace:

• messaging.{system}.message.*: Describes attributes for individual messages
• messaging.{system}.destination.*: Describes the destination a message (or a batch) are published to and received from respectively. The combination of attributes in these namespaces should uniquely identify the entity and include properties significant for this messaging system. For example, Kafka instrumentations should include partition identifier.
• messaging.{system}.consumer.*: Describes message consumer properties
• messaging.{system}.batch.*: Describes message batch properties

Examples

Topic with multiple consumers

Given is a publisher that publishes a message to a topic exchange “T” on RabbitMQ, and two consumers which both get the message delivered.

flowchart LR;
  subgraph PRODUCER
  direction TB
  P[Span Publish A]
  end
  subgraph CONSUMER1
  direction TB
  R1[Span Deliver A 1]
  end
  subgraph CONSUMER2
  direction TB
  R2[Span Deliver A 2]
  end
  P-. link .-R1;
  P-. link .-R2;

  classDef normal fill:green
  class P,R1,R2 normal
  linkStyle 0,1 color:green,stroke:green

Batch receiving

Given is a publisher that publishes two messages to a topic “Q” on Kafka, and a consumer which receives both messages in one batch.

flowchart LR;
  subgraph PRODUCER
  direction TB
  PA[Span Publish A]
  PB[Span Publish B]
  end
  subgraph CONSUMER1
  direction TB
  D1[Span Receive A B]
  end
  PA-. link .-D1;
  PB-. link .-D1;

  classDef normal fill:green
  class PA,PB,D1 normal
  linkStyle 0,1 color:green,stroke:green

Semantic Conventions for specific messaging technologies

More specific Semantic Conventions are defined for the following messaging technologies:

• Kafka: Semantic Conventions for Apache Kafka.
• RabbitMQ: Semantic Conventions for RabbitMQ.
• RocketMQ: Semantic Conventions for Apache RocketMQ.