Skip to content

Latest commit

 

History

History
256 lines (195 loc) · 16.1 KB

File metadata and controls

256 lines (195 loc) · 16.1 KB

Semantic conventions for HTTP spans

This document defines semantic conventions for HTTP client and server Spans. They can be used for http and https schemes and various HTTP versions like 1.1, 2 and SPDY.

Name

HTTP spans MUST follow the overall guidelines for span names. Many REST APIs encode parameters into URI path, e.g. /api/users/123 where 123 is a user id, which creates high cardinality value space not suitable for span names. In case of HTTP servers, these endpoints are often mapped by the server frameworks to more concise HTTP routes, e.g. /api/users/{user_id}, which are recommended as the low cardinality span names. However, the same approach usually does not work for HTTP client spans, especially when instrumentation is provided by a lower-level middleware that is not aware of the specifics of how the URIs are formed. Therefore, HTTP client spans SHOULD be using conservative, low cardinality names formed from the available parameters of an HTTP request, such as "HTTP {METHOD_NAME}". Instrumentation MUST NOT default to using URI path as span name, but MAY provide hooks to allow custom logic to override the default span name.

Status

Implementations MUST set the span status if the HTTP communication failed or an HTTP error status code is returned (e.g. above 3xx).

In the case of an HTTP redirect, the request should normally be considered successful, unless the client aborts following redirects due to hitting some limit (redirect loop). If following a (chain of) redirect(s) successfully, the status should be set according to the result of the final HTTP request.

Don't set the span status description if the reason can be inferred from http.status_code and http.status_text.

HTTP code Span status code
100...299 Ok
3xx redirect codes DeadlineExceeded in case of loop (see above) [1], otherwise Ok
401 Unauthorized ⚠ Unauthenticated ⚠ (Unauthorized actually means unauthenticated according to RFC 7235)
403 Forbidden PermissionDenied
404 Not Found NotFound
429 Too Many Requests ResourceExhausted
Other 4xx code InvalidArgument [1]
501 Not Implemented Unimplemented
503 Service Unavailable Unavailable
504 Gateway Timeout DeadlineExceeded
Other 5xx code Internal [1]
Any status code the client fails to interpret (e.g., 093 or 573) Unknown

Note that the items marked with [1] are different from the mapping defined in the OpenCensus semantic conventions.

Common Attributes

Attribute name Notes and examples Required?
http.method HTTP request method. E.g. "GET". Yes
http.url Full HTTP request URL in the form scheme://host[:port]/path?query[#fragment]. Usually the fragment is not transmitted over HTTP, but if it is known, it should be included nevertheless. Defined later.
http.target The full request target as passed in a HTTP request line or equivalent, e.g. "/path/12314/?q=ddds#123". Defined later.
http.host The value of the HTTP host header. When the header is empty or not present, this attribute should be the same. Defined later.
http.scheme The URI scheme identifying the used protocol: "http" or "https" Defined later.
http.status_code HTTP response status code. E.g. 200 (integer) If and only if one was received/sent.
http.status_text HTTP reason phrase. E.g. "OK" No
http.flavor Kind of HTTP protocol used: "1.0", "1.1", "2", "SPDY" or "QUIC". No
http.user_agent Value of the HTTP User-Agent header sent by the client. No
http.request_content_length The size of the request payload body in bytes. This is the number of bytes transferred excluding headers and is often, but not always, present as the Content-Length header. For requests using transport encoding, this should be the compressed size. No
http.request_content_length_uncompressed The size of the uncompressed request payload body after transport decoding. Not set if transport encoding not used. No
http.response_content_length The size of the response payload body in bytes. This is the number of bytes transferred excluding headers and is often, but not always, present as the Content-Length header. For requests using transport encoding, this should be the compressed size. No
http.response_content_length_uncompressed The size of the uncompressed response payload body after transport decoding. Not set if transport encoding not used. No

It is recommended to also use the general network attributes, especially net.peer.ip. If net.transport is not specified, it can be assumed to be IP.TCP except if http.flavor is QUIC, in which case IP.UDP is assumed.

HTTP client

This span type represents an outbound HTTP request.

For an HTTP client span, SpanKind MUST be Client.

If set, http.url must be the originally requested URL, before any HTTP-redirects that may happen when executing the request.

One of the following sets of attributes is required (in order of usual preference unless for a particular web client/framework it is known that some other set is preferable for some reason; all strings must be non-empty):

  • http.url
  • http.scheme, http.host, http.target
  • http.scheme, net.peer.name, net.peer.port, http.target
  • http.scheme, net.peer.ip, net.peer.port, http.target

Note that in some cases http.host might be different from the net.peer.name used to look up the net.peer.ip that is actually connected to. In that case it is strongly recommended to set the net.peer.name attribute in addition to http.host.

For status, the following special cases have canonical error codes assigned:

Client error Trace status code
DNS resolution failed Unknown
Request cancelled by caller Cancelled
URL cannot be parsed InvalidArgument
Request timed out DeadlineExceeded

This is not meant to be an exhaustive list but if there is no clear mapping for some error conditions, instrumentation developers are encouraged to use Unknown and open a PR or issue in the specification repository.

HTTP server

To understand the attributes defined in this section, it is helpful to read the "Definitions" subsection.

HTTP server definitions

This section gives a short summary of some concepts in web server configuration and web app deployment that are relevant to tracing.

Usually, on a physical host, reachable by one or multiple IP addresses, a single HTTP listener process runs. If multiple processes are running, they must listen on distinct TCP/UDP ports so that the OS can route incoming TCP/UDP packets to the right one.

Within a single server process, there can be multiple virtual hosts. The HTTP host header (in combination with a port number) is normally used to determine to which of them to route incoming HTTP requests.

The host header value that matches some virtual host is called the virtual hosts's server name. If there are multiple aliases for the virtual host, one of them (often the first one listed in the configuration) is called the primary server name. See for example, the Apache ServerName or NGINX server_name directive or the CGI specification on SERVER_NAME (RFC 3875). In practice the HTTP host header is often ignored when just a single virtual host is configured for the IP.

Within a single virtual host, some servers support the concepts of an HTTP application (for example in Java, the Servlet JSR defines an application as "a collection of servlets, HTML pages, classes, and other resources that make up a complete application on a Web server" -- SRV.9 in JSR 53; in a deployment of a Python application to Apache, the application would be the PEP 3333 conformant callable that is configured using the WSGIScriptAlias directive of mod_wsgi).

An application can be "mounted" under some application root (also know as context root context prefix, or document base) which is a fixed path prefix of the URL that determines to which application a request is routed (e.g., the server could be configured to route all requests that go to an URL path starting with /webshop/ at a particular virtual host to the com.example.webshop web application).

Some servers allow to bind the same HTTP application to multiple (virtual host, application root) pairs.

TODO: Find way to trace HTTP application and application root (opentelemetry/opentelementry-specification#335)

HTTP Server semantic conventions

This span type represents an inbound HTTP request.

For an HTTP server span, SpanKind MUST be Server.

Given an inbound request for a route (e.g. "/users/:userID?") the name attribute of the span SHOULD be set to this route. If the route does not include the application root, it SHOULD be prepended to the span name.

If the route cannot be determined, the name attribute MUST be set as defined in the general semantic conventions for HTTP.

Attribute name Notes and examples Required?
http.server_name The primary server name of the matched virtual host. This should be obtained via configuration. If no such configuration can be obtained, this attribute MUST NOT be set ( net.host.name should be used instead). [1]
http.route The matched route (path template). (TODO: Define whether to prepend application root) E.g. "/users/:userID?". No
http.client_ip The IP address of the original client behind all proxies, if known (e.g. from X-Forwarded-For). Note that this is not necessarily the same as net.peer.ip, which would identify the network-level peer, which may be a proxy. No

[1]: http.url is usually not readily available on the server side but would have to be assembled in a cumbersome and sometimes lossy process from other information (see e.g. open-telemetry/opentelemetry-python#148). It is thus preferred to supply the raw data that is available. Namely, one of the following sets is required (in order of usual preference unless for a particular web server/framework it is known that some other set is preferable for some reason; all strings must be non-empty):

  • http.scheme, http.host, http.target
  • http.scheme, http.server_name, net.host.port, http.target
  • http.scheme, net.host.name, net.host.port, http.target
  • http.url

Of course, more than the required attributes can be supplied, but this is recommended only if they cannot be inferred from the sent ones. For example, http.server_name has shown great value in practice, as bogus HTTP Host headers occur often in the wild.

It is strongly recommended to set http.server_name to allow associating requests with some logical server entity.

HTTP client-server example

As an example, if a browser request for https://example.com:8080/webshop/articles/4?s=1 is invoked from a host with IP 192.0.2.4, we may have the following Span on the client side:

Span name: /webshop/articles/4 (NOTE: This is subject to change, see open-telemetry/opentelemetry-specification#270.)

Attribute name Value
http.method "GET"
http.flavor "1.1"
http.url "https://example.com:8080/webshop/articles/4?s=1"
net.peer.ip "192.0.2.5"
http.status_code 200
http.status_text "OK"

The corresponding server Span may look like this:

Span name: /webshop/articles/:article_id.

Attribute name Value
http.method "GET"
http.flavor "1.1"
http.target "/webshop/articles/4?s=1"
http.host "example.com:8080"
http.server_name "example.com"
net.host.port 8080
http.scheme "https"
http.route "/webshop/articles/:article_id"
http.status_code 200
http.status_text "OK"
http.client_ip "192.0.2.4"
net.peer.ip "192.0.2.5" (the client goes through a proxy)
http.user_agent "Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:72.0) Gecko/20100101 Firefox/72.0"

Note that following the recommendations above, http.url is not set in the above example. If set, it would be "https://example.com:8080/webshop/articles/4?s=1" but due to http.scheme, http.host and http.target being set, it would be redundant. As explained above, these separate values are preferred but if for some reason the URL is available but the other values are not, URL can replace http.scheme, http.host and http.target.