HTTP header manipulation

The HTTP connection manager manipulates several HTTP headers both during decoding (when the request is being received) as well as during encoding (when the response is being sent).

:scheme

Envoy will always set the :scheme header while processing a request. It should always be available to filters, and should be forwarded upstream for HTTP/2 and HTTP/3, where x-forwarded-proto will be sent for HTTP/1.1.

For HTTP/2, and HTTP/3, incoming :scheme headers are trusted and propogated through upstream. For HTTP/1, the :scheme header will be set 1) From the absolute URL if present and valid. An invalid (not “http” or “https”) scheme, or an https scheme over an unencrypted connection will result in Envoy rejecting the request. This is the only scheme validation Envoy performs as it avoids a HTTP/1.1-specific privilege escalation attack for edge Envoys [1] which doesn’t have a comparable vector for HTTP/2 and above [2]. 2) From the value of the x-forwarded-proto header after sanitization (to valid x-forwarded-proto from trusted downstreams, otherwise based on downstream encryption level).

This default behavior can be overridden via the scheme_header_transformation configuration option.

The :scheme header will be used by Envoy over x-forwarded-proto where the URI scheme is wanted, for example serving content from cache based on the :scheme header rather than X-Forwarded-Proto, or setting the scheme of redirects based on the scheme of the original URI. See Why is Envoy operating on X-Forwarded-Proto instead of :scheme or vice-versa? for more details.

[1]

Edge Envoys often have plaintext HTTP/1.1 listeners. If Envoy trusts absolute URL scheme from fully qualfied URLs, a MiTM can adjust relative URLs to https absolute URLs, and inadvertently cause the Envoy’s upstream to send PII or other sensitive data over what it then believes is a secure connection.

[2]

Unlike HTTP/1.1, HTTP/2 is in practice always served over TLS via ALPN for edge Envoys. In mesh networks using insecure HTTP/2, if the downstream is not trusted to set scheme, the scheme_header_transformation should be used.

:path

The :path header is a pseudo-header populated by Envoy using the value of the path of the HTTP request. E.g. an HTTP request of the form GET /docs/thing HTTP/1.1 would have a :path value of /docs/thing.

:method

The :method header is a pseudo-header populated by Envoy using the value of the method of the HTTP request. E.g. an HTTP request of the form GET /docs/thing HTTP/1.1 would have a :method value of GET. Values are in upper-case, e.g. GET, PUT, POST, and DELETE. Other possible values are described in HTTP Methods <https://developer.mozilla.org/en-US/docs/Web/HTTP/Methods>.

user-agent

The user-agent header may be set by the connection manager during decoding if the add_user_agent option is enabled. The header is only modified if it is not already set. If the connection manager does set the header, the value is determined by the --service-cluster command line option.

server

The server header will be set during encoding to the value in the server_name option.

referer

The referer header will be sanitized during decoding. Multiple URLs, invalid relative URLs containing a fragment component, and valid absolute URLs containing userinfo or a fragment component will be removed.

x-client-trace-id

If an external client sets this header, Envoy will join the provided trace ID with the internally generated x-request-id. x-client-trace-id needs to be globally unique and generating a uuid4 is recommended. If this header is set, it has similar effect to x-envoy-force-trace. See the tracing.client_enabled runtime configuration setting.

x-envoy-downstream-service-cluster

Internal services often want to know which service is calling them. This header is cleaned from external requests, but for internal requests will contain the service cluster of the caller. Note that in the current implementation, this should be considered a hint as it is set by the caller and could be easily spoofed by any internal entity. In the future Envoy will support a mutual authentication TLS mesh which will make this header fully secure. Like user-agent, the value is determined by the --service-cluster command line option. In order to enable this feature you need to set the user_agent option to true.

x-envoy-downstream-service-node

Internal services may want to know the downstream node request comes from. This header is quite similar to x-envoy-downstream-service-cluster, except the value is taken from the --service-node option.

x-envoy-external-address

It is a common case where a service wants to perform analytics based on the origin client’s IP address. Per the lengthy discussion on XFF, this can get quite complicated, so Envoy simplifies this by setting x-envoy-external-address to the trusted client address if the request is from an external client. x-envoy-external-address is not set or overwritten for internal requests. This header can be safely forwarded between internal services for analytics purposes without having to deal with the complexities of XFF.

x-envoy-force-trace

If an internal request sets this header, Envoy will modify the generated x-request-id such that it forces traces to be collected. This also forces x-request-id to be returned in the response headers. If this request ID is then propagated to other hosts, traces will also be collected on those hosts which will provide a consistent trace for an entire request flow. See the tracing.global_enabled and tracing.random_sampling runtime configuration settings.

x-envoy-internal

It is a common case where a service wants to know whether a request is internal origin or not. Envoy uses XFF to determine this and then will set the header value to true.

This is a convenience to avoid having to parse and understand XFF.

x-envoy-original-dst-host

The header used to override destination address when using the Original Destination load balancing policy.

It is ignored, unless the use of it is enabled via use_http_header.

x-forwarded-client-cert

x-forwarded-client-cert (XFCC) is a proxy header which indicates certificate information of part or all of the clients or proxies that a request has flowed through, on its way from the client to the server. A proxy may choose to sanitize/append/forward the XFCC header before proxying the request.

The XFCC header value is a comma (“,”) separated string. Each substring is an XFCC element, which holds information added by a single proxy. A proxy can append the current client certificate information as an XFCC element, to the end of the request’s XFCC header after a comma.

Each XFCC element is a semicolon “;” separated string. Each substring is a key-value pair, grouped together by an equals (“=”) sign. The keys are case-insensitive, the values are case-sensitive. If “,”, “;” or “=” appear in a value, the value should be double-quoted. Double-quotes in the value should be replaced by backslash-double-quote (").

The following keys are supported:

1. By The Subject Alternative Name (URI type) of the current proxy’s certificate. The current proxy’s certificate may contain multiple URI type Subject Alternative Names, each will be a separate key-value pair.

2. Hash The SHA 256 digest of the current client certificate.

3. Cert The entire client certificate in URL encoded PEM format.

4. Chain The entire client certificate chain (including the leaf certificate) in URL encoded PEM format.

5. Subject The Subject field of the current client certificate. The value is always double-quoted.

6. URI The URI type Subject Alternative Name field of the current client certificate. A client certificate may contain multiple URI type Subject Alternative Names, each will be a separate key-value pair.

7. DNS The DNS type Subject Alternative Name field of the current client certificate. A client certificate may contain multiple DNS type Subject Alternative Names, each will be a separate key-value pair.

A client certificate may contain multiple Subject Alternative Name types. For details on different Subject Alternative Name types, please refer RFC 2459.

Some examples of the XFCC header are:

1. For one client certificate with only URI type Subject Alternative Name: x-forwarded-client-cert: By=http://frontend.lyft.com;Hash=468ed33be74eee6556d90c0149c1309e9ba61d6425303443c0748a02dd8de688;Subject="/C=US/ST=CA/L=San Francisco/OU=Lyft/CN=Test Client";URI=http://testclient.lyft.com

2. For two client certificates with only URI type Subject Alternative Name: x-forwarded-client-cert: By=http://frontend.lyft.com;Hash=468ed33be74eee6556d90c0149c1309e9ba61d6425303443c0748a02dd8de688;URI=http://testclient.lyft.com,By=http://backend.lyft.com;Hash=9ba61d6425303443c0748a02dd8de688468ed33be74eee6556d90c0149c1309e;URI=http://frontend.lyft.com

3. For one client certificate with both URI type and DNS type Subject Alternative Name: x-forwarded-client-cert: By=http://frontend.lyft.com;Hash=468ed33be74eee6556d90c0149c1309e9ba61d6425303443c0748a02dd8de688;Subject="/C=US/ST=CA/L=San Francisco/OU=Lyft/CN=Test Client";URI=http://testclient.lyft.com;DNS=lyft.com;DNS=www.lyft.com

How Envoy processes XFCC is specified by the forward_client_cert_details and the set_current_client_cert_details HTTP connection manager options. If forward_client_cert_details is unset, the XFCC header will be sanitized by default.

x-forwarded-for

x-forwarded-for (XFF) is a standard proxy header which indicates the IP addresses that a request has flowed through on its way from the client to the server. A compliant proxy will append the IP address of the nearest client to the XFF list before proxying the request. Some examples of XFF are:

1. x-forwarded-for: 50.0.0.1 (single client)

2. x-forwarded-for: 50.0.0.1, 40.0.0.1 (external proxy hop)

3. x-forwarded-for: 50.0.0.1, 10.0.0.1 (internal proxy hop)

Envoy will only append to XFF if the use_remote_address HTTP connection manager option is set to true and the skip_xff_append is set false. This means that if use_remote_address is false (which is the default) or skip_xff_append is true, the connection manager operates in a transparent mode where it does not modify XFF.

Attention

In general, use_remote_address should be set to true when Envoy is deployed as an edge node (aka a front proxy), whereas it may need to be set to false when Envoy is used as an internal service node in a mesh deployment.

The value of use_remote_address controls how Envoy determines the trusted client address. Given an HTTP request that has traveled through a series of zero or more proxies to reach Envoy, the trusted client address is the earliest source IP address that is known to be accurate. The source IP address of the immediate downstream node’s connection to Envoy is trusted. XFF sometimes can be trusted. Malicious clients can forge XFF, but the last address in XFF can be trusted if it was put there by a trusted proxy.

Alternatively, Envoy supports extensions for determining the trusted client address or original IP address.

Note

The use of such extensions cannot be mixed with use_remote_address nor xff_num_trusted_hops.

Envoy’s default rules for determining the trusted client address (before appending anything to XFF) are:

• If use_remote_address is false and an XFF containing at least one IP address is present in the request, the trusted client address is the last (rightmost) IP address in XFF.

• Otherwise, the trusted client address is the source IP address of the immediate downstream node’s connection to Envoy.

In an environment where there are one or more trusted proxies in front of an edge Envoy instance, the xff_num_trusted_hops configuration option can be used to trust additional addresses from XFF:

• If use_remote_address is false and xff_num_trusted_hops is set to a value N that is greater than zero, the trusted client address is the (N+1)th address from the right end of XFF. (If the XFF contains fewer than N+1 addresses, Envoy falls back to using the immediate downstream connection’s source address as trusted client address.)

• If use_remote_address is true and xff_num_trusted_hops is set to a value N that is greater than zero, the trusted client address is the Nth address from the right end of XFF. (If the XFF contains fewer than N addresses, Envoy falls back to using the immediate downstream connection’s source address as trusted client address.)

Envoy uses the trusted client address contents to determine whether a request originated externally or internally. This influences whether the x-envoy-internal header is set.

Example 1: Envoy as edge proxy, without a trusted proxy in front of it

Settings:

use_remote_address = true

xff_num_trusted_hops = 0

Request details:

Downstream IP address = 192.0.2.5

XFF = “203.0.113.128, 203.0.113.10, 203.0.113.1”

Result:

Trusted client address = 192.0.2.5 (XFF is ignored)

X-Envoy-External-Address is set to 192.0.2.5

XFF is changed to “203.0.113.128, 203.0.113.10, 203.0.113.1, 192.0.2.5”

X-Envoy-Internal is removed (if it was present in the incoming request)

Example 2: Envoy as internal proxy, with the Envoy edge proxy from Example 1 in front of it

Settings:

use_remote_address = false

xff_num_trusted_hops = 0

Request details:

Downstream IP address = 10.11.12.13 (address of the Envoy edge proxy)

XFF = “203.0.113.128, 203.0.113.10, 203.0.113.1, 192.0.2.5”

Result:

Trusted client address = 192.0.2.5 (last address in XFF is trusted)

X-Envoy-External-Address is not modified

X-Envoy-Internal is removed (if it was present in the incoming request)

Example 3: Envoy as edge proxy, with two trusted external proxies in front of it

Settings:

use_remote_address = true

xff_num_trusted_hops = 2

Request details:

Downstream IP address = 192.0.2.5

XFF = “203.0.113.128, 203.0.113.10, 203.0.113.1”

Result:

Trusted client address = 203.0.113.10 (2nd to last address in XFF is trusted)

X-Envoy-External-Address is set to 203.0.113.10

XFF is changed to “203.0.113.128, 203.0.113.10, 203.0.113.1, 192.0.2.5”

X-Envoy-Internal is removed (if it was present in the incoming request)

Example 4: Envoy as internal proxy, with the edge proxy from Example 3 in front of it

Settings:

use_remote_address = false

xff_num_trusted_hops = 2

Request details:

Downstream IP address = 10.11.12.13 (address of the Envoy edge proxy)

XFF = “203.0.113.128, 203.0.113.10, 203.0.113.1, 192.0.2.5”

Result:

Trusted client address = 203.0.113.10

X-Envoy-External-Address is not modified

X-Envoy-Internal is removed (if it was present in the incoming request)

Example 5: Envoy as an internal proxy, receiving a request from an internal client

Settings:

use_remote_address = false

xff_num_trusted_hops = 0

Request details:

Downstream IP address = 10.20.30.40 (address of the internal client)

XFF is not present

Result:

Trusted client address = 10.20.30.40

X-Envoy-External-Address remains unset

X-Envoy-Internal is set to “false”

Example 6: The internal Envoy from Example 5, receiving a request proxied by another Envoy

Settings:

use_remote_address = false

xff_num_trusted_hops = 0

Request details:

Downstream IP address = 10.20.30.50 (address of the Envoy instance proxying to this one)

XFF = “10.20.30.40”

Result:

Trusted client address = 10.20.30.40

X-Envoy-External-Address remains unset

X-Envoy-Internal is set to “true”

A few very important notes about XFF:

1. If use_remote_address is set to true, Envoy sets the x-envoy-external-address header to the trusted client address.

2. XFF is what Envoy uses to determine whether a request is internal origin or external origin. If use_remote_address is set to true, the request is internal if and only if the request contains no XFF and the immediate downstream node’s connection to Envoy has an internal (RFC1918 or RFC4193) source address. If use_remote_address is false, the request is internal if and only if XFF contains a single RFC1918 or RFC4193 address.

   • NOTE: If an internal service proxies an external request to another internal service, and includes the original XFF header, Envoy will append to it on egress if use_remote_address is set. This will cause the other side to think the request is external. Generally, this is what is intended if XFF is being forwarded. If it is not intended, do not forward XFF, and forward x-envoy-internal instead.

   • NOTE: If an internal service call is forwarded to another internal service (preserving XFF), Envoy will not consider it internal. This is a known “bug” due to the simplification of how XFF is parsed to determine if a request is internal. In this scenario, do not forward XFF and allow Envoy to generate a new one with a single internal origin IP.

x-forwarded-host

The x-forwarded-host header is a de-facto standard proxy header which indicates the original host requested by the client in the :authority (host in HTTP1) header. A compliant proxy appends the original value of the :authority header to x-forwarded-host only if the :authority header is modified.

Envoy updates the :authority header if a host rewrite option (one of host_rewrite_literal, auto_host_rewrite, host_rewrite_header, or host_rewrite_path_regex) is used and appends its original value to x-forwarded-host if append_x_forwarded_host is set.

x-forwarded-port

Usually, the x-forwarded-port header comes with the x-forwarded-proto header for service to know the originating destination port of the connection, which is the listener port in Envoy’s side.

Envoy will append the x-forwarded-port header if append_x_forwarded_port is set to true and the header has not been set.

Downstream x-forwarded-port headers will only be trusted if xff_num_trusted_hops is non-zero. If xff_num_trusted_hops is zero, downstream x-forwarded-port headers will be overwritten.

x-forwarded-proto

It is a common case where a service wants to know what the originating protocol (HTTP or HTTPS) was of the connection terminated by front/edge Envoy. x-forwarded-proto contains this information. It will be set to either http or https.

Downstream x-forwarded-proto headers will only be trusted if xff_num_trusted_hops is non-zero. If xff_num_trusted_hops is zero, downstream x-forwarded-proto headers and :scheme headers will be set to http or https based on if the downstream connection is TLS or not.

If the scheme is changed via the scheme_header_transformation configuration option, x-forwarded-proto will be updated as well.

The x-forwarded-proto header will be used by Envoy over :scheme where the underlying encryption is wanted, for example clearing default ports based on x-forwarded-proto. See Why is Envoy operating on X-Forwarded-Proto instead of :scheme or vice-versa? for more details.

x-request-id

The x-request-id header is used by Envoy to uniquely identify a request as well as perform stable access logging and tracing. Envoy will generate an x-request-id header for all external origin requests (the header is sanitized). It will also generate an x-request-id header for internal requests that do not already have one. This means that x-request-id can and should be propagated between client applications in order to have stable IDs across the entire mesh. Due to the out of process architecture of Envoy, the header can not be automatically forwarded by Envoy itself. This is one of the few areas where a thin client library is needed to perform this duty. How that is done is out of scope for this documentation. If x-request-id is propagated across all hosts, the following features are available:

• Stable access logging via the v3 API runtime filter.

• Stable tracing when performing random sampling via the tracing.random_sampling runtime setting or via forced tracing using the x-envoy-force-trace and x-client-trace-id headers.

See the architecture overview on context propagation for more information.

x-ot-span-context

The x-ot-span-context HTTP header is used by Envoy to establish proper parent-child relationships between tracing spans when used with the LightStep tracer. For example, an egress span is a child of an ingress span (if the ingress span was present). Envoy injects the x-ot-span-context header on ingress requests and forwards it to the local service. Envoy relies on the application to propagate x-ot-span-context on the egress call to an upstream. See more on tracing here.

x-b3-traceid

The x-b3-traceid HTTP header is used by the Zipkin tracer in Envoy. The TraceId is 64-bit in length and indicates the overall ID of the trace. Every span in a trace shares this ID. See more on zipkin tracing here.

x-b3-spanid

The x-b3-spanid HTTP header is used by the Zipkin tracer in Envoy. The SpanId is 64-bit in length and indicates the position of the current operation in the trace tree. The value should not be interpreted: it may or may not be derived from the value of the TraceId. See more on zipkin tracing here.

x-b3-parentspanid

The x-b3-parentspanid HTTP header is used by the Zipkin tracer in Envoy. The ParentSpanId is 64-bit in length and indicates the position of the parent operation in the trace tree. When the span is the root of the trace tree, the ParentSpanId is absent. See more on zipkin tracing here.

x-b3-sampled

The x-b3-sampled HTTP header is used by the Zipkin tracer in Envoy. When the Sampled flag is either not specified or set to 1, the span will be reported to the tracing system. Once Sampled is set to 0 or 1, the same value should be consistently sent downstream. See more on zipkin tracing here.

x-b3-flags

The x-b3-flags HTTP header is used by the Zipkin tracer in Envoy. The encode one or more options. For example, Debug is encoded as X-B3-Flags: 1. See more on zipkin tracing here.

b3

The b3 HTTP header is used by the Zipkin tracer in Envoy. Is a more compressed header format. See more on zipkin tracing here.

x-datadog-trace-id

The x-datadog-trace-id HTTP header is used by the Datadog tracer in Envoy. The 64-bit value represents the ID of the overall trace, and is used to correlate the spans.

x-datadog-parent-id

The x-datadog-parent-id HTTP header is used by the Datadog tracer in Envoy. The 64-bit value uniquely identifies the span within the trace, and is used to create parent-child relationships between spans.

x-datadog-sampling-priority

The x-datadog-sampling-priority HTTP header is used by the Datadog tracer in Envoy. The integer value indicates the sampling decision that has been made for this trace. A value of 0 indicates that the trace should not be collected, and a value of 1 requests that spans are sampled and reported.

sw8

The sw8 HTTP header is used by the SkyWalking tracer in Envoy. It contains the key tracing context for the SkyWalking tracer and is used to establish the relationship between the tracing spans of downstream and Envoy. See more on SkyWalking tracing here.

x-amzn-trace-id

The x-amzn-trace-id HTTP header is used by the AWS X-Ray tracer in Envoy. The trace ID, parent ID and sampling decision are added to HTTP requests in the tracing header. See more on AWS X-Ray tracing here.

Custom request/response headers

Custom request/response headers can be added to a request/response at the weighted cluster, route, virtual host, and/or global route configuration level. See the v3 API documentation.

Neither :-prefixed pseudo-headers nor the Host: header may be modified via this mechanism. The :path and :authority headers may instead be modified via mechanisms such as prefix_rewrite, regex_rewrite, and host_rewrite.

Headers are appended to requests/responses in the following order: weighted cluster level headers, route level headers, virtual host level headers and finally global level headers.

Envoy supports adding dynamic values to request and response headers. The percent symbol (%) is used to delimit variable names.

Attention

If a literal percent symbol (%) is desired in a request/response header, it must be escaped by doubling it. For example, to emit a header with the value 100%, the custom header value in the Envoy configuration must be 100%%.

All HTTP Command Operators used for access logging may be specified in custom request/response headers. However, depending where a particular command operator is used, the context needed for the operator may not be available and the produced output is empty string. For example, the following configuration uses %RESPONSE_CODE% operator to modify request headers using code from the response. The output is an empty string, because request headers are modified before the request is sent upstream and the response is not received yet.

header_formatters.yaml

          route_config:
            name: local_route
            request_headers_to_add:
            - header:
                key: "response-code"
                value: "%RESPONSE_CODE%"

Attention

The following legacy header formatters are still supported, but will be deprecated in the future. The equivalent information can be accessed using indicated substitutes.

%DYNAMIC_METADATA(["namespace", "key", ...])%

Populates the header with dynamic metadata available in a request (e.g.: added by filters like the header-to-metadata filter).

This works both on request and response headers.

Use %DYNAMIC_METADATA(namespace:key:…):Z% instead.

%UPSTREAM_METADATA(["namespace", "key", ...])%

Populates the header with EDS endpoint metadata from the upstream host selected by the router. Metadata may be selected from any namespace. In general, metadata values may be strings, numbers, booleans, lists, nested structures, or null. Upstream metadata values may be selected from nested structs by specifying multiple keys. Otherwise, only string, boolean, and numeric values are supported. If the namespace or key(s) are not found, or if the selected value is not a supported type, then no header is emitted. The namespace and key(s) are specified as a JSON array of strings. Finally, percent symbols in the parameters do not need to be escaped by doubling them.

Upstream metadata cannot be added to request headers as the upstream host has not been selected when custom request headers are generated.

Use %UPSTREAM_METADATA(namespace:key:…):Z% instead.

%PER_REQUEST_STATE(reverse.dns.data.name)%

Populates the header with values set on the stream info filterState() object. To be usable in custom request/response headers, these values must be of type Envoy::Router::StringAccessor. These values should be named in standard reverse DNS style, identifying the organization that created the value and ending in a unique name for the data.

Use %FILTER_STATE(reverse.dns.data.name:PLAIN):Z% instead.