Semantic Conventions for Kafka

Status: Experimental

The Semantic Conventions for Apache Kafka extend and override the Messaging Semantic Conventions that describe common messaging operations attributes in addition to the Semantic Conventions described on this page.

messaging.system MUST be set to "kafka".

Span attributes

For Apache Kafka, the following additional attributes are defined:

Attribute	Type	Description	Examples	Requirement Level	Stability
messaging.operation.type	string	A string identifying the type of the messaging operation. [1]	publish; create; receive	Required
error.type	string	Describes a class of error the operation ended with. [2]	amqp:decode-error; KAFKA_STORAGE_ERROR; channel-error	Conditionally Required If and only if the messaging operation has failed.
messaging.batch.message_count	int	The number of messages sent, received, or processed in the scope of the batching operation. [3]	0; 1; 2	Conditionally Required [4]
messaging.destination.name	string	The message destination name [5]	MyQueue; MyTopic	Conditionally Required [6]
messaging.kafka.message.tombstone	boolean	A boolean that is true if the message is a tombstone.		Conditionally Required [7]
server.address	string	Server domain name if available without reverse DNS lookup; otherwise, IP address or Unix domain socket name. [8]	example.com; 10.1.2.80; /tmp/my.sock	Conditionally Required If available.
messaging.client.id	string	A unique identifier for the client that consumes or produces a message.	client-5; myhost@8742@s8083jm	Recommended
messaging.destination.partition.id	string	String representation of the partition id the message (or batch) is sent to or received from.	1	Recommended
messaging.kafka.consumer.group	string	Name of the Kafka Consumer Group that is handling the message. Only applies to consumers, not producers.	my-group	Recommended
messaging.kafka.message.key	string	Message keys in Kafka are used for grouping alike messages to ensure they’re processed on the same partition. They differ from messaging.message.id in that they’re not unique. If the key is null, the attribute MUST NOT be set. [9]	myKey	Recommended If span describes operation on a single message.
messaging.kafka.message.offset	int	The offset of a record in the corresponding Kafka partition.	42	Recommended If span describes operation on a single message.
messaging.message.body.size	int	The size of the message body in bytes. [10]	1439	Recommended If span describes operation on a single message.
messaging.message.id	string	A value used by the messaging system as an identifier for the message, represented as a string.	452a7c7c7c7048c2f887f61572b18fc2	Recommended If span describes operation on a single message.
messaging.operation.name	string	The system-specific name of the messaging operation.	ack; nack; send	Recommended [11]
server.port	int	Server port number. [12]	80; 8080; 443	Recommended

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

[2]: The error.type SHOULD be predictable, and SHOULD have low cardinality.

When error.type is set to a type (e.g., an exception type), its canonical class name identifying the type within the artifact SHOULD be used.

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.

[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]: 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.

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

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

[8]: Server domain name of the broker if available without reverse DNS lookup; otherwise, IP address or Unix domain socket name.

[9]: If the key type is not string, it’s string representation has to be supplied for the attribute. If the key has no unambiguous, canonical string form, don’t include its value.

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

[11]: If the operation is not sufficiently described by messaging.operation.type.

[12]: 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.

Value	Description	Stability
_OTHER	A fallback error value to be used when the instrumentation doesn’t define a custom value.

messaging.operation.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.

Value	Description	Stability
publish	One or more messages are provided for publishing to an intermediary. If a single message is published, the context of the “Publish” span can be used as the creation context and no “Create” span needs to be created.
create	A message is created. “Create” spans always refer to a single message and are used to provide a unique creation context for messages in batch publishing scenarios.
receive	One or more messages are requested by a consumer. This operation refers to pull-based scenarios, where consumers explicitly call methods of messaging SDKs to receive messages.
process	One or more messages are delivered to or processed by a consumer.
settle	One or more messages are settled.

For Apache Kafka producers, peer.service SHOULD be set to the name of the broker or service the message will be sent to. The service.name of a Consumer’s Resource SHOULD match the peer.service of the Producer, when the message is directly passed to another service. If an intermediary broker is present, service.name and peer.service will not be the same.

messaging.client.id SHOULD be set to the client name of a consumer or producer, which is unique for each individual instance.

Examples

Apache Kafka with Quarkus or Spring Boot Example

Given is a process P, that publishes a message to a topic T1 on Apache Kafka. One process, CA, receives the message and publishes a new message to a topic T2 that is then received and processed by CB.

Frameworks such as Quarkus and Spring Boot separate processing of a received message from producing subsequent messages out. For this reason, receiving (Span Rcv1) is the parent of both processing (Span Proc1) and producing a new message (Span Prod2). The span representing message receiving (Span Rcv1) should not set messaging.operation.type to receive, as it does not only receive the message but also converts the input message to something suitable for the processing operation to consume and creates the output message from the result of processing.

Process P:  | Span Prod1 |
--
Process CA:              | Span Rcv1 |
                                | Span Proc1 |
                                  | Span Prod2 |
--
Process CB:                           | Span Rcv2 |