Create custom headers in backend services

This page describes how to configure custom headers in backend services used by the classic Application Load Balancer.

Custom request and response headers let you specify additional headers that the load balancer can add to HTTP(S) requests and responses. Depending on the information detected by the load balancer, these headers can include the following information:

  • Latency to the client
  • Geographic location of the client's IP address
  • Parameters of the TLS connection

Custom request headers are supported for backend services, while custom response headers are supported for backend services and backend buckets.

The load balancer adds certain headers by default to all HTTP(S) requests and responses that it proxies between backends and clients. For more information, see Target proxies.

Before you begin

  • If necessary, update to the latest version of the Google Cloud CLI:

    gcloud components update
    

How custom headers work

Custom headers work as follows:

  • When the global external Application Load Balancer makes a request to the backend, the load balancer adds request headers.

  • The global external Application Load Balancer sets response headers before returning the responses to the client.

To enable custom headers, you specify a list of headers in a property of the backend service or backend bucket.

You specify each header as a header-name:header-value string. The header must contain a colon separating the header name and header value.

Header names must fulfill the following requirements:

  • The header name must be a valid HTTP header-field name definition (per RFC 7230).
  • The header name must not be X-User-IP or CDN-Loop.
  • The following hop-by-hop headers must not be used: Keep-Alive, Transfer-Encoding, TE, Connection, Trailer, Upgrade, Proxy-Authorization, and Proxy-Authenticate. In accordance with RFC 2616, these headers are not stored by caches or propagated by the target proxies.
  • The header name must not begin with X-Google, X-Goog-, X-GFE or X-Amz-.
  • A header name must not appear more than once in the list of added headers.

Header values must fulfill the following requirements:

  • The header value must be a valid HTTP header field definition (per RFC 7230, with obsolete forms disallowed).
  • The header value can be blank.
  • The header value can include one or more variables, enclosed by curly braces, that expand to values that the load balancer provides. A list of variables allowed in the header value is in the next section.

For example, you might specify a header with two variable names, for the client region and client city. The gcloud command-line tool has a flag for specifying request headers, --custom-request-header. Make sure to use only straight quotation marks (') with this flag.

The general format for the flag is:

    --custom-request-header='HEADER_NAME:[HEADER_VALUE]'

Any variables must be enclosed in curly braces. For example:

    --custom-request-header 'X-Client-Geo-Location:{client_region},{client_city}'

For clients located in Mountain View, California, the load balancer adds a header as follows:

X-Client-Geo-Location:US,Mountain View

Variables supported with header values

The following variables can appear in custom header values.

Variable Description
cdn_cache_id Location code and ID of the cache instance used to serve the request. This is the same value populated in the jsonPayload.cacheId field of the Cloud CDN request logs in Logging.
cdn_cache_status Current cache status. Values can be hit, miss, revalidated, stale, uncacheable, or disabled for any object served by a Cloud CDN-enabled backend.
origin_request_header Reflects the value of the Origin header in the request for Cross-Origin Resource Sharing (CORS) use cases.
client_rtt_msec Estimated round-trip transmission time between the load balancer and the HTTP(S) client, in milliseconds. This is the smoothed round-trip time (SRTT) parameter measured by the load balancer's TCP stack, per RFC 2988. Smoothed RTT is an algorithm that deals with variations and anomalies that may occur in RTT measurements.
client_region The country (or region) associated with the client's IP address. This is a Unicode CLDR region code, such as US or FR. (For most countries, these codes correspond directly to ISO-3166-2 codes.)
client_region_subdivision Subdivision, e.g. a province or state, of the country associated with the client's IP address. This is a Unicode CLDR subdivision ID, such as USCA or CAON. (These Unicode codes are derived from the subdivisions defined by the ISO-3166-2 standard.)
client_city Name of the city from which the request originated, for example, Mountain View for Mountain View, California. There is no canonical list of valid values for this variable. The city names can contain US-ASCII letters, numbers, spaces, and the following characters: !#$%&'*+-.^_`|~.
client_city_lat_long Latitude and Longitude of the city from which the request originated, for example, 37.386051,-122.083851 for a request from Mountain View.
client_ip_address The client's IP address. This is the same as the client IP address that is the next-to-last address in the X-Forwarded-For header.
client_port The client's source port.
client_encrypted true if the connection between the client and the load balancer is encrypted (using HTTPS, HTTP/2 or HTTP/3); otherwise, false.
client_protocol The HTTP protocol used for communication between the client and the load balancer. One of HTTP/1.0, HTTP/1.1, HTTP/2, or HTTP/3.
server_ip_address The IP address of the load balancer that the client connects to. This can be useful when multiple load balancers share common backends. This is the same as the last IP address in the X-Forwarded-For header.
server_port The destination port number that the client connects to.
tls_sni_hostname Server name indication (as defined in RFC 6066), if provided by the client during the TLS or QUIC handshake. The hostname is converted to lowercase and with any trailing dot removed.
tls_version TLS version negotiated between client and load balancer during the SSL handshake. Possible values include: TLSv1, TLSv1.1, TLSv1.2, and TLSv1.3. If the client connects using QUIC instead of TLS, the value is QUIC.
tls_cipher_suite Cipher suite negotiated during the TLS handshake. The value is four hex digits defined by the IANA TLS Cipher Suite Registry, for example, 009C for TLS_RSA_WITH_AES_128_GCM_SHA256. This value is empty for QUIC and for unencrypted client connections.
tls_ja3_fingerprint JA3 TLS/SSL fingerprint if the client connects using HTTPS, HTTP/2 or HTTP/3.

The load balancer expands variables to empty strings when it cannot determine their values. For example:

  • Geographic location variables when the IP address's location is unknown
  • TLS parameters when TLS is not in use
  • The {origin_request_header} when the request does not include an Origin header
  • The {cdn_cache_status} header when included in a request header

Geographic values (regions, subdivisions, and cities) are estimates based on the client's IP address. From time to time, Google updates the data that provides these values in order to improve accuracy and to reflect geographic and political changes.

Headers added by the load balancer overwrite any existing headers that have the same name. Header names are case insensitive. When header names are passed to an HTTP/2 backend, the HTTP/2 protocol encodes header names as lower case.

In header values, leading whitespace and trailing whitespace are insignificant, and are not passed to the backend. To allow for curly braces in header values, the load balancer interprets two opening curly braces ({{) as a single opening brace ({), and two closing curly braces (}}) as a single closing brace (}).

Mutual TLS custom headers

The following additional header variables are available if mutual TLS (mTLS) is configured on the load balancer's TargetHttpsProxy.

Variable Description
client_cert_present true if the client has provided a certificate during the TLS handshake; otherwise, false.
client_cert_chain_verified true if the client certificate chain is verified against a configured TrustStore; otherwise, false.
client_cert_error Predefined strings representing the error conditions. For more information about the error strings, see mTLS client validation modes.
client_cert_sha256_fingerprint Base64-encoded SHA-256 fingerprint of the client certificate.
client_cert_serial_number The serial number of the client certificate. If the serial number is longer than 50 bytes, the client_cert_error is set to client_cert_serial_number_exceeded_size_limit, and the serial number is set to an empty string.
client_cert_spiffe_id

The SPIFFE ID from the Subject Alternative Name (SAN) field. If the value is not valid or exceeds 2048 bytes, the SPIFFE ID is set to an empty string.

If the SPIFFE ID is longer than 2048 bytes, the client_cert_error is set to client_cert_spiffe_id_exceeded_size_limit.

client_cert_uri_sans

Comma-separated Base64-encoded list of the SAN extensions of type URI. The SAN extensions are extracted from the client certificate. The SPIFFE ID is not included in the client_cert_uri_sans field.

If the client_cert_uri_sans is longer than 512 bytes, the client_cert_error is set to client_cert_uri_sans_exceeded_size_limit, and the comma-separated list is set to an empty string.

client_cert_dnsname_sans

Comma-separated Base64-encoded list of the SAN extensions of type DNSName. The SAN extensions are extracted from the client certificate.

If the client_cert_dnsname_sans is longer than 512 bytes, the client_cert_error is set to client_cert_dnsname_sans_exceeded_size_limit, and the comma-separated list is set to an empty string.

client_cert_valid_not_before Timestamp (RFC 3339 date string format) before which the client certificate is not valid. For example, 2022-07-01T18:05:09+00:00.
client_cert_valid_not_after Timestamp (RFC 3339 date string format) after which the client certificate is not valid. For example, 2022-07-01T18:05:09+00:00.
client_cert_issuer_dn

Base64-encoded DER encoding of the full Issuer field from the certificate.

If the client_cert_issuer_dn is longer than 512 bytes, the string client_cert_issuer_dn_exceeded_size_limit is added to client_cert_error, and client_cert_issuer_dn is set to an empty string.

client_cert_subject_dn

Base64-encoded DER encoding of the full Subject field from the certificate.

If the client_cert_subject_dn is longer than 512 bytes, the string client_cert_subject_dn_exceeded_size_limit is added to client_cert_error, and client_cert_subject_dn is set to an empty string.

client_cert_leaf

The client leaf certificate for an established mTLS connection where the certificate passed validation. Certificate encoding is compliant with RFC 9440. This means the binary DER certificate is encoded using Base64 and delimited with colons on either side.

If client_cert_leaf exceeds 16 KB unencoded, the string client_cert_validated_leaf_exceeded_size_limit is added to client_cert_error, and client_cert_leaf is set to an empty string.

client_cert_chain

The comma-delimited list of certificates, in standard TLS order, of the client certificate chain for an established mTLS connection where the client certificate passed validation, not including the leaf certificate. Certificate encoding is compliant with RFC 9440.

If the combined size of client_cert_leaf and client_cert_chain before Base64 encoding exceeds 16 KB, the string client_cert_validated_chain_exceeded_size_limit is added to client_cert_error, and client_cert_chain is set to an empty string.

Configure custom request headers

Console

To add custom request headers to an existing backend service:

  1. Go to the load balancing summary page.
    Go to the Load balancing page
  2. Click Backends.
  3. Click the name of a backend service.
  4. Click Edit .
  5. Click Advanced configurations (Session affinity, connection draining timeout, security policies).
  6. Under Custom request headers, click Add header.
  7. Enter the Header name and Header value for the custom request header.
  8. Enter any additional custom request headers.
  9. Click Save.

To remove a custom request header from a backend service:

  1. Go to the load balancing summary page.
    Go to the Load balancing page
  2. Click Backends.
  3. Click the name of a backend service.
  4. Click Edit .
  5. Click Advanced configurations (Session affinity, connection draining timeout, security policies).
  6. Click the X next to the name of the custom request header you want to remove.
  7. Click Save.

gcloud

To specify custom request headers, use the gcloud compute backend-services create or gcloud compute backend-services update command with the --custom-request-header flag.

To create a backend service with custom request headers:

gcloud compute backend-services create BACKEND_SERVICE_NAME \
    --global-health-checks \
    --global \
    --protocol HTTPS \
    --health-checks https-basic-check \
    --custom-request-header='HEADER_NAME:[HEADER_VALUE]'

To add more request headers, specify a unique header name and value by repeating the --custom-request-header flag.

To add custom headers to an existing backend service:

gcloud compute backend-services update BACKEND_SERVICE_NAME \
    --global \
    --custom-request-header='HEADER_1_NAME:[HEADER_1_VALUE]' \
    --custom-request-header='HEADER_2_NAME:[HEADER_2_VALUE]'

The previous step replaces any headers already in the backend service with the request headers that you specify in the command.

To remove all headers from a backend service:

gcloud compute backend-services update BACKEND_SERVICE_NAME \
    --global \
    --no-custom-request-headers

API

Make a PATCH request to the backendServices.patch method:

PATCH https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/global/backendServices/BACKEND_SERVICE_NAME
"customRequestHeaders": [
   "client_city:Mountain View"
]

Terraform

For a Terraform script that creates a load balancer with custom headers, see Terraform examples for global external Application Load Balancers.

Configure custom response headers

Console

To add custom response headers to an existing backend service:

  1. Go to the load balancing summary page.
    Go to the Load balancing page
  2. Click Backends.
  3. Click the name of a backend service.
  4. Click Edit .
  5. Click Advanced configurations (Session affinity, connection draining timeout, security policies).
  6. Under Custom response headers, click Add header.
  7. Enter the Header name and Header value for the custom response header.
  8. Enter any additional custom response headers.
  9. Click Save.

To remove a custom response header from a backend service:

  1. Go to the load balancing summary page.
    Go to the Load balancing page
  2. Click Backends.
  3. Click the name of a backend service.
  4. Click Edit .
  5. Click Advanced configurations (Session affinity, connection draining timeout, security policies).
  6. Click the X next to the name of the custom response header you want to remove.
  7. Click Save.

gcloud

For backend services, use the gcloud compute backend-services create or gcloud compute backend-services update command with the --custom-response-header flag.

For backend buckets, use the gcloud compute backend-buckets create or gcloud compute backend-buckets update command with the --custom-response-header flag.

gcloud compute backend-services (create | update) BACKEND_SERVICE_NAME
    --custom-response-header='HEADER_NAME:[HEADER_VALUE]'
gcloud compute backend-buckets (create | update) BACKEND_BUCKET_NAME
    --custom-response-header='HEADER_NAME:[HEADER_VALUE]'

Example with X-Frame-Options header:

gcloud compute backend-buckets update gaming-lab \
    --custom-response-header='X-Frame-Options: DENY'

Example with Strict-Transport-Security header:

The following example shows you how to add a custom response header to support HTTP Strict Transport Security (HSTS):

gcloud compute backend-services update customer-bs-name \
    --global \
    --custom-response-header='Strict-Transport-Security: max-age=63072000'

API

For backend buckets, use the Method: backendBuckets.insert or Method: backendBuckets.update API call.

For backend services, use the Method: backendServices.insert or Method: backendServices.update API call.

Use one of the following API calls:

POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/global/backendBuckets

PUT https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/global/backendBuckets/BACKEND_BUCKET_NAME

POST https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/global/backendServices

PUT https://compute.googleapis.com/compute/v1/projects/PROJECT_ID/global/backendServices/BACKEND_SERVICE_NAME

Add the following snippet to the JSON request body:

"customResponseHeaders":HEADER_NAME:[HEADER_VALUE]

Terraform

For a Terraform script that creates a load balancer with custom headers, see Terraform examples for global external Application Load Balancers.

Set response headers for Cloud Storage

If you need to set HTTP headers on responses from Cloud Storage—such as Cross-Origin Resource Policy, X-Frame-Options, or X-XSS-Protection headers—Google Cloud offers the option to use custom headers for Cloud CDN with Cloud Storage. You can do this by configuring custom headers at the load balancer backend bucket level, as described on this page.

Custom response headers configured at the backend bucket level are only added to responses if the client request is sent to the load balancer IP address. Custom headers are not added to responses if the client's request was made directly to the Cloud Storage API.

Use custom headers with Google Cloud Armor

When you configure a Google Cloud Armor security policy, you can configure Google Cloud Armor to insert a custom header and value. If your Google Cloud Armor security policy is configured to insert the same custom header name as your global external Application Load Balancer or classic Application Load Balancer's custom headers, then the header value specified in your Google Cloud Armor security policy is overwritten by the value populated by the load balancer. If you don't want the Google Cloud Armor policy to be overwritten, ensure that you don't use the same name.

Limitations

The following limitations apply to custom headers used with global load balancers:

  • You can specify a maximum of 16 custom request headers for each backend service.
  • You can specify a maximum of 16 custom response headers for each backend service.
  • The total size of all custom request headers per backend service (name and value combined, before variable expansion) cannot exceed 8 KB.
  • The total size of all custom response headers per backend service (name and value combined, before variable expansion) cannot exceed 8 KB.

What's next