Access logging

Configuration

Access logs are configured as part of the HTTP connection manager config or TCP Proxy.

Format Rules

Access log formats contain command operators that extract the relevant data and insert it. They support two formats: “format strings” and “format dictionaries”. In both cases, the command operators are used to extract the relevant data, which is then inserted into the specified log format. Only one access log format may be specified at a time.

Format Strings

Format strings are plain strings, specified using the format key. They may contain either command operators or other characters interpreted as a plain string. The access log formatter does not make any assumptions about a new line separator, so one has to specified as part of the format string. See the default format for an example.

Default Format String

If custom format string is not specified, Envoy uses the following default format:

[%START_TIME%] "%REQ(:METHOD)% %REQ(X-ENVOY-ORIGINAL-PATH?:PATH)% %PROTOCOL%"
%RESPONSE_CODE% %RESPONSE_FLAGS% %BYTES_RECEIVED% %BYTES_SENT% %DURATION%
%RESP(X-ENVOY-UPSTREAM-SERVICE-TIME)% "%REQ(X-FORWARDED-FOR)%" "%REQ(USER-AGENT)%"
"%REQ(X-REQUEST-ID)%" "%REQ(:AUTHORITY)%" "%UPSTREAM_HOST%"\n

Example of the default Envoy access log format:

[2016-04-15T20:17:00.310Z] "POST /api/v1/locations HTTP/2" 204 - 154 0 226 100 "10.0.35.28"
"nsq2http" "cc21d9b0-cf5c-432b-8c7e-98aeb7988cd2" "locations" "tcp://10.0.2.1:80"

Format Dictionaries

Format dictionaries are dictionaries that specify a structured access log output format, specified using the json_format key. This allows logs to be output in a structured format such as JSON. Similar to format strings, command operators are evaluated and their values inserted into the format dictionary to construct the log output.

For example, with the following format provided in the configuration:

{
  "config": {
    "json_format": {
        "protocol": "%PROTOCOL%",
        "duration": "%DURATION%",
        "my_custom_header": "%REQ(MY_CUSTOM_HEADER)%"
    }
  }
}

The following JSON object would be written to the log file:

{"protocol": "HTTP/1.1", "duration": "123", "my_custom_header": "value_of_MY_CUSTOM_HEADER"}

This allows you to specify a custom key for each command operator.

Format dictionaries have the following restrictions:

  • The dictionary must map strings to strings (specifically, strings to command operators). Nesting is not currently supported.

Command Operators

Command operators are used to extract values that will be inserted into the access logs. The same operators are used by different types of access logs (such as HTTP and TCP). Some fields may have slightly different meanings, depending on what type of log it is. Differences are noted.

Note that if a value is not set/empty, the logs will contain a ‘-‘ character.

The following command operators are supported:

%START_TIME%
HTTP
Request start time including milliseconds.
TCP
Downstream connection start time including milliseconds.

START_TIME can be customized using a format string. In addition to that, START_TIME also accepts following specifiers:

Specifier Explanation
%s The number of seconds since the Epoch
%f, %[1-9]f Fractional seconds digits, default is 9 digits (nanosecond)
  • %3f millisecond (3 digits)
  • %6f microsecond (6 digits)
  • %9f nanosecond (9 digits)

Examples of formatting START_TIME is as follows:

%START_TIME(%Y/%m/%dT%H:%M:%S%z %s)%

# To include millisecond fraction of the second (.000 ... .999). E.g. 1527590590.528.
%START_TIME(%s.%3f)%

%START_TIME(%s.%6f)%

%START_TIME(%s.%9f)%
%BYTES_RECEIVED%
HTTP
Body bytes received.
TCP
Downstream bytes received on connection.
%PROTOCOL%
HTTP
Protocol. Currently either HTTP/1.1 or HTTP/2.
TCP
Not implemented (“-“).
%RESPONSE_CODE%
HTTP
HTTP response code. Note that a response code of ‘0’ means that the server never sent the beginning of a response. This generally means that the (downstream) client disconnected.
TCP
Not implemented (“-“).
%BYTES_SENT%
HTTP
Body bytes sent. For WebSocket connection it will also include response header bytes.
TCP
Downstream bytes sent on connection.
%DURATION%
HTTP
Total duration in milliseconds of the request from the start time to the last byte out.
TCP
Total duration in milliseconds of the downstream connection.
%RESPONSE_DURATION%
HTTP
Total duration in milliseconds of the request from the start time to the first byte read from the upstream host.
TCP
Not implemented (“-“).
%RESPONSE_FLAGS%

Additional details about the response or connection, if any. For TCP connections, the response codes mentioned in the descriptions do not apply. Possible values are:

HTTP and TCP
  • UH: No healthy upstream hosts in upstream cluster in addition to 503 response code.
  • UF: Upstream connection failure in addition to 503 response code.
  • UO: Upstream overflow (circuit breaking) in addition to 503 response code.
  • NR: No route configured for a given request in addition to 404 response code.
HTTP only
  • LH: Local service failed health check request in addition to 503 response code.
  • UT: Upstream request timeout in addition to 504 response code.
  • LR: Connection local reset in addition to 503 response code.
  • UR: Upstream remote reset in addition to 503 response code.
  • UC: Upstream connection termination in addition to 503 response code.
  • DI: The request processing was delayed for a period specified via fault injection.
  • FI: The request was aborted with a response code specified via fault injection.
  • RL: The request was ratelimited locally by the HTTP rate limit filter in addition to 429 response code.
%RESPONSE_TX_DURATION%
HTTP
Total duration in milliseconds of the request from the first byte read from the upstream host to the last byte sent downstream.
TCP
Not implemented (“-“).
%UPSTREAM_HOST%
Upstream host URL (e.g., tcp://ip:port for TCP connections).
%UPSTREAM_CLUSTER%
Upstream cluster to which the upstream host belongs to.
%UPSTREAM_LOCAL_ADDRESS%
Local address of the upstream connection. If the address is an IP address it includes both address and port.
%DOWNSTREAM_REMOTE_ADDRESS%

Remote address of the downstream connection. If the address is an IP address it includes both address and port.

Note

This may not be the physical remote address of the peer if the address has been inferred from proxy proto or x-forwarded-for.

%DOWNSTREAM_REMOTE_ADDRESS_WITHOUT_PORT%

Remote address of the downstream connection. If the address is an IP address the output does not include port.

Note

This may not be the physical remote address of the peer if the address has been inferred from proxy proto or x-forwarded-for.

%DOWNSTREAM_LOCAL_ADDRESS%
Local address of the downstream connection. If the address is an IP address it includes both address and port. If the original connection was redirected by iptables REDIRECT, this represents the original destination address restored by the Original Destination Filter using SO_ORIGINAL_DST socket option. If the original connection was redirected by iptables TPROXY, and the listener’s transparent option was set to true, this represents the original destination address and port.
%DOWNSTREAM_LOCAL_ADDRESS_WITHOUT_PORT%
Same as %DOWNSTREAM_LOCAL_ADDRESS% excluding port if the address is an IP address.
%REQ(X?Y):Z%
HTTP
An HTTP request header where X is the main HTTP header, Y is the alternative one, and Z is an optional parameter denoting string truncation up to Z characters long. The value is taken from the HTTP request header named X first and if it’s not set, then request header Y is used. If none of the headers are present ‘-‘ symbol will be in the log.
TCP
Not implemented (“-“).
%RESP(X?Y):Z%
HTTP
Same as %REQ(X?Y):Z% but taken from HTTP response headers.
TCP
Not implemented (“-“).
%TRAILER(X?Y):Z%
HTTP
Same as %REQ(X?Y):Z% but taken from HTTP response trailers.
TCP
Not implemented (“-“).
%DYNAMIC_METADATA(NAMESPACE:KEY*):Z%
HTTP

Dynamic Metadata info, where NAMESPACE is the the filter namespace used when setting the metadata, KEY is an optional lookup up key in the namespace with the option of specifying nested keys separated by ‘:’, and Z is an optional parameter denoting string truncation up to Z characters long. Dynamic Metadata can be set by filters using the StreamInfo API: setDynamicMetadata. The data will be logged as a JSON string. For example, for the following dynamic metadata:

com.test.my_filter: {"test_key": "foo", "test_object": {"inner_key": "bar"}}

  • %DYNAMIC_METADATA(com.test.my_filter)% will log: {"test_key": "foo", "test_object": {"inner_key": "bar"}}
  • %DYNAMIC_METADATA(com.test.my_filter:test_key)% will log: "foo"
  • %DYNAMIC_METADATA(com.test.my_filter:test_object)% will log: {"inner_key": "bar"}
  • %DYNAMIC_METADATA(com.test.my_filter:test_object:inner_key)% will log: "bar"
  • %DYNAMIC_METADATA(com.unknown_filter)% will log: -
  • %DYNAMIC_METADATA(com.test.my_filter:unknown_key)% will log: -
  • %DYNAMIC_METADATA(com.test.my_filter):25% will log (truncation at 25 characters): {"test_key": "foo", "test
TCP
Not implemented (“-“).
%REQUESTED_SERVER_NAME%
HTTP
String value set on ssl connection socket for Server Name Indication (SNI)
TCP
String value set on ssl connection socket for Server Name Indication (SNI)