Table of Contents
Cross-cutting concerns send their state to the next process using
Propagator
s, which are defined as objects used to read and write
context data to and from messages exchanged by the applications.
Each concern creates a set of Propagator
s for every supported
Propagator
type.
Propagator
s leverage the Context
to inject and extract data for each
cross-cutting concern, such as traces and Baggage
.
Propagation is usually implemented via a cooperation of library-specific request
interceptors and Propagators
, where the interceptors detect incoming and outgoing requests and use the Propagator
's extract and inject operations respectively.
The Propagators API is expected to be leveraged by users writing instrumentation libraries.
A Propagator
type defines the restrictions imposed by a specific transport
and is bound to a data type, in order to propagate in-band context data across process boundaries.
The Propagators API currently defines one Propagator
type:
TextMapPropagator
is a type that inject values into and extracts values from carriers as string key/value pairs.
A binary Propagator
type will be added in the future (see #437).
A carrier is the medium used by Propagator
s to read values from and write values to.
Each specific Propagator
type defines its expected carrier type, such as a string map
or a byte array.
Carriers used at Inject are expected to be mutable.
Propagator
s MUST define Inject
and Extract
operations, in order to write
values to and read values from carriers respectively. Each Propagator
type MUST define the specific carrier type
and MAY define additional parameters.
Injects the value into a carrier. For example, into the headers of an HTTP request.
Required arguments:
- A
Context
. The Propagator MUST retrieve the appropriate value from theContext
first, such asSpanContext
,Baggage
or another cross-cutting concern context. - The carrier that holds the propagation fields. For example, an outgoing message or HTTP request.
Extracts the value from an incoming request. For example, from the headers of an HTTP request.
If a value can not be parsed from the carrier for a cross-cutting concern,
the implementation MUST NOT throw an exception and MUST NOT store a new value in the Context
,
in order to preserve any previously existing valid value.
Required arguments:
- A
Context
. - The carrier that holds the propagation fields. For example, an incoming message or http response.
Returns a new Context
derived from the Context
passed as argument,
containing the extracted value, which can be a SpanContext
,
Baggage
or another cross-cutting concern context.
TextMapPropagator
performs the injection and extraction of a cross-cutting concern
value as string key/values pairs into carriers that travel in-band across process boundaries.
The carrier of propagated data on both the client (injector) and server (extractor) side is usually an HTTP request.
In order to increase compatibility, the key/value pairs MUST only consist of US-ASCII characters that make up valid HTTP header fields as per RFC 7230.
Getter
and Setter
are optional helper components used for extraction and injection respectively,
and are defined as separate objects from the carrier to avoid runtime allocations,
by removing the need for additional interface-implementing-objects wrapping the carrier in order
to access its contents.
Getter
and Setter
MUST be stateless and allowed to be saved as constants, in order to effectively
avoid runtime allocations.
The predefined propagation fields. If your carrier is reused, you should delete the fields here before calling inject.
Fields are defined as string keys identifying format-specific components in a carrier.
For example, if the carrier is a single-use or immutable request object, you don't need to clear fields as they couldn't have been set before. If it is a mutable, retryable object, successive calls should clear these fields first.
The use cases of this are:
- allow pre-allocation of fields, especially in systems like gRPC Metadata
- allow a single-pass over an iterator
Returns list of fields that will be used by the TextMapPropagator
.
Observe that some Propagator
s may define, besides the returned values, additional fields with
variable names. To get a full list of fields for a specific carrier object, use the
Keys operation.
Injects the value into a carrier. The required arguments are the same as defined by the base Inject operation.
Optional arguments:
- A
Setter
to set a propagation key/value pair. Propagators MAY invoke it multiple times in order to set multiple pairs. This is an additional argument that languages are free to define to help inject data into the carrier.
Setter is an argument in Inject
that sets values into given fields.
Setter
allows a TextMapPropagator
to set propagated fields into a carrier.
One of the ways to implement it is Setter
class with Set
method as described below.
Replaces a propagated field with the given value.
Required arguments:
- the carrier holding the propagation fields. For example, an outgoing message or an HTTP request.
- the key of the field.
- the value of the field.
The implementation SHOULD preserve casing (e.g. it should not transform Content-Type
to content-type
) if the used protocol is case insensitive, otherwise it MUST preserve casing.
Extracts the value from an incoming request. The required arguments are the same as defined by the base Extract operation.
Optional arguments:
- A
Getter
invoked for each propagation key to get. This is an additional argument that languages are free to define to help extract data from the carrier.
Returns a new Context
derived from the Context
passed as argument.
Getter is an argument in Extract
that get value from given field
Getter
allows a TextMapPropagator
to read propagated fields from a carrier.
One of the ways to implement it is Getter
class with Get
and Keys
methods
as described below. Languages may decide on alternative implementations and
expose corresponding methods as delegates or other ways.
The Keys
function MUST return the list of all the keys in the carrier.
Required arguments:
- The carrier of the propagation fields, such as an HTTP request.
The Keys
function can be called by Propagator
s which are using variable key names in order to
iterate over all the keys in the specified carrier.
For example, it can be used to detect all keys following the uberctx-{user-defined-key}
pattern, as defined by the
Jaeger Propagation Format.
The Get function MUST return the first value of the given propagation key or return null if the key doesn't exist.
Required arguments:
- the carrier of propagation fields, such as an HTTP request.
- the key of the field.
The Get function is responsible for handling case sensitivity. If the getter is intended to work with a HTTP request object, the getter MUST be case insensitive.
Languages can choose to implement a Propagator
type as a single object
exposing Inject
and Extract
methods, or they can opt to divide the
responsibilities further into individual Injector
s and Extractor
s. A
Propagator
can be implemented by composing individual Injector
s and
Extractors
.
Implementations MUST offer a facility to group multiple Propagator
s
from different cross-cutting concerns in order to leverage them as a
single entity.
A composite propagator can be built from a list of propagators, or a list of
injectors and extractors. The resulting composite Propagator
will invoke the Propagator
s, Injector
s, or Extractor
s, in the order they were specified.
Each composite Propagator
will implement a specific Propagator
type, such
as TextMapPropagator
, as different Propagator
types will likely operate on different
data types.
There MUST be functions to accomplish the following operations.
- Create a composite propagator
- Extract from a composite propagator
- Inject into a composite propagator
Required arguments:
- A list of
Propagator
s or a list ofInjector
s andExtractor
s.
Returns a new composite Propagator
with the specified Propagator
s.
Required arguments:
- A
Context
. - The carrier that holds propagation fields.
- The instance of
Getter
invoked for each propagation key to get.
Required arguments:
- A
Context
. - The carrier that holds propagation fields.
- The
Setter
to set a propagation key/value pair. Propagators MAY invoke it multiple times in order to set multiple pairs.
The OpenTelemetry API MUST provide a way to obtain a propagator for each
supported Propagator
type. Instrumentation libraries SHOULD call propagators
to extract and inject the context on all remote calls. Propagators, depending on
the language, MAY be set up using various dependency injection techniques or
available as global accessors.
Note: It is a discouraged practice, but certain instrumentation libraries might use proprietary context propagation protocols or be hardcoded to use a specific one. In such cases, instrumentation libraries MAY choose not to use the API-provided propagators and instead hardcode the context extraction and injection logic.
The OpenTelemetry API MUST use no-op propagators unless explicitly configured otherwise. Context propagation may be used for various telemetry signals - traces, metrics, logging and more. Therefore, context propagation MAY be enabled for any of them independently. For instance, a span exporter may be left unconfigured, although the trace context propagation was configured to enrich logs or metrics.
Platforms such as ASP.NET may pre-configure out-of-the-box
propagators. If pre-configured, Propagator
s SHOULD default to a composite
Propagator
containing the W3C Trace Context Propagator and the Baggage
Propagator
specified in the Baggage API.
These platforms MUST also allow pre-configured propagators to be disabled or overridden.
This method MUST exist for each supported Propagator
type.
Returns a global Propagator
. This usually will be composite instance.
This method MUST exist for each supported Propagator
type.
Sets the global Propagator
instance.
Required parameters:
- A
Propagator
. This usually will be a composite instance.
The official list of propagators that MUST be maintained by the OpenTelemetry organization and MUST be distributed as OpenTelemetry extension packages:
- W3C TraceContext. MAY alternatively be distributed as part of the OpenTelemetry API.
- W3C Baggage. MAY alternatively be distributed as part of the OpenTelemetry API.
- B3.
- Jaeger.
Additional Propagator
s implementing vendor-specific protocols such as AWS
X-Ray (Note, AWS is used as an example, not as a requirement) trace header
protocol can be either maintained and distributed by their respective vendors or
as part of the OpenTelemetry organization. The reasons for maintaining those as
a community are:
- Propagators are small pieces of code and their functionality is often publicly documented (unlike exporters).
- People will often need to use propagators that are not specific to their tracing or metrics vendor. For example, customers of tracing vendor may still want to use an Cloud vendor-specific propagator for requests to the services of this cloud vendor.
- Only a small number of propagators will need to exist, and this number will shrink as vendors and users shift to W3C TraceContext.
B3 has both single and multi-header encodings. It also has semantics that do not map directly to OpenTelemetry such as a debug trace flag, and allowing spans from both sides of request to share the same id. To maximize compatibility between OpenTelemetry and Zipkin implementations, the following guidelines have been established for B3 context propagation.
When extracting B3, propagators:
- MUST attempt to extract B3 encoded using single and multi-header formats. The single-header variant takes precedence over the multi-header version.
- MUST preserve a debug trace flag, if received, and propagate it with subsequent requests. Additionally, an OpenTelemetry implementation MUST set the sampled trace flag when the debug flag is set.
- MUST NOT reuse
X-B3-SpanId
as the id for the server-side span.
When injecting B3, propagators:
- MUST default to injecting B3 using the single-header format
- MUST provide configuration to change the default injection format to B3 multi-header
- MUST NOT propagate
X-B3-ParentSpanId
as OpenTelemetry does not support reusing the same id for both sides of a request.