Sending HTTP headers to subgraphs
Configure which headers the Apollo Router sends to which subgraphs
You can configure which HTTP headers the Apollo Router includes in its requests to each of your subgraphs. You can define per-subgraph header rules, along with rules that apply to all subgraphs.
You define header rules in your YAML configuration file, like so:
# ...other configuration...headers:all: # Header rules for all subgraphsrequest:- propagate:matching: ^upstream-header-.*- remove:named: "x-legacy-account-id"subgraphs:products: # Header rules for just the products subgraphrequest:- insert:name: "router-subgraph-name"value: "products"
Supported header rules
The Apollo Router supports the following types of header rules:
propagate
Enables you to selectively pass along headers that were included in the client's request to the router.
You can specify which headers to propagate based on a matching regex pattern:
- propagate:matching: .*
Note: The Apollo Router never propagates so-called hop-by-hop headers (such as Content-Length
) when propagating by pattern.
Alternatively, you can provide a static string via the named
option. These named
configurations have additional flexibility, because they support the following options:
default
: A value to set if no value was sent by the clientrename
: Renames the header's key to the provided value
- propagate:named: "x-user-id"default: "abc123"rename: "account-id"
remove
Enables you to selectively remove headers that were included in the client's request to the router. Like propagate
, this option can match either a static string or a regular expression.
# Do not send this subgraph the "Cookie" header.- remove:named: "Cookie"- remove:# Remove headers that include the legacy 'x-' prefix.matching: ^x-.*$
insert
Enables you to add custom headers to requests going to a specific subgraph. These headers are strings (statics, coming from request body or from context) that originate in the router, instead of originating in the client.
- Insert static header
- insert:name: "sent-from-our-apollo-router"value: "indeed"
- Insert header from context
- insert:name: "sent-from-our-apollo-router-context"from_context: "my_key_in_context"
- Insert header from request body
- insert:name: "sent-from-our-apollo-router-request-body"path: ".operationName" # It's a JSON path query to fetch the operation name from request bodydefault: "UNKNOWN" # If no operationName has been specified
Example JSON path queries
Let's say you have a JSON request body with the following structure:
{"query": "{ products { id name } }","extensions": {"metadata": [{"app_name": "random_app_name"}]}}
To fetch the value of the field app_name
, the corresponding path is .extensions.metadata[0].app_name
.
JSON path queries always begin with a period .
With this configuration:
headers:all:request:- insert:name: from_app_namepath: .extensions.metadata[0].app_name
You will pass a header to all your subgraphs: "from_app_name": "random_app_name"
Rule ordering
Header rules are applied in the same order they're declared, and later rules can override the effects of earlier rules. Consider this example:
❌
headers:all:request:- remove:named: "test"- propagate:matching: .*
In this example, first any header named test
is removed from the list of headers to propagate. However, the list of headers to propagate is currently empty! Next, the propagate
rule adds all headers to the propagation list, including test
.
To correctly remove a header from the propagation list, make sure to define your remove
rule after any propagate
rules:
✅
headers:all:request:- propagate:matching: .*- remove:named: "test"
With this ordering, first all headers are added to the propagation list, then the test
header is removed.
Example
Here's a complete example showing all the possible configuration options in use:
headers:# Header rules for all subgraphsall:request:# Propagate matching headers- propagate:matching: ^upstream-header-.*# Propagate matching headers- propagate:named: "some-header"default: "default-value"rename: "destination-header"# Remove the "x-legacy-account-id" header- remove:named: "x-legacy-account-id"# Remove matching headers- remove:matching: ^x-deprecated-.*# Insert the 'my-company' header- insert:name: "my-company"value: "acme"# Subgraph-specific header rulessubgraphs:products:request:# Calls to the products subgraph have the "router-subgraph-name" header set to `products`.- insert:name: "router-subgraph-name"value: "products"accounts:request:# Calls to the accounts subgraph have the "router-subgraph-name" header set to `accounts`.- insert:name: "router-subgraph-name"value: "accounts"
Response header propagation
It is not currently possible to propagate response headers from subgraphs to clients using YAML configuration alone. However, you can achieve this using Rhai scripting.
This approach relies on the fact that each request has a context
object that can store data for the duration of that request:
- For each subgraph response, copy header values into context.
- For the supergraph response, copy header values from the context onto the response.
Example router.yaml
that will use the Rhai script:
rhai:main: "main.rhai"
Example Rhai script that collects set-cookie
headers from subgraphs and merges them into a single client response header:
fn supergraph_service(service) {let add_cookies_to_response = |response| {if response.context["set_cookie_headers"]?.len > 0 {response.headers["set-cookie"] = response.context["set_cookie_headers"];}};service.map_response(add_cookies_to_response);}fn subgraph_service(service, subgraph) {let store_cookies_from_subgraphs = |response| {if "set-cookie" in response.headers {if response.context["set_cookie_headers"] == () {response.context.set_cookie_headers = []}response.context.set_cookie_headers += response.headers.values("set-cookie");}};service.map_response(store_cookies_from_subgraphs);}
If you require a configuration-based solution for response header propagation, please leave a comment on our issue tracker.
Propagation between subgraphs
It is not currently possible to propagate headers between subgraphs using YAML config alone. However, you can achieve this using Rhai scripting.
This approach relies on the fact that each request has a context
object that can store data for the duration of that request:
- For each subgraph response, copy header values into context.
- For each subgraph request, copy header values from context into the subgraph request.
Example router.yaml
that will use the Rhai script:
rhai:main: "main.rhai"
Example Rhai script that copies request-id
and user
headers:
fn subgraph_service(service, subgraph) {// The list of headers that you which to propagate.let headers = ["request-id", "user"];// Callback for subgraph requests. Inserts headers from context into the subgraph request.let request_callback = |request| {for key in headers {if request.context[key] != () {request.subgraph.headers[key] = request.context[key];}}};// Callback for subgraph responses. Pulls header values out of the response and inserts them into context.let response_callback = |response| {for key in headers {if key in response.headers {response.context[key] = response.headers[key];}}};// Register the callbacks.service.map_request(request_callback);service.map_response(response_callback);}
If you require a configuration-based solution for response header propagation, please leave a comment on our issue tracker.