Harness Proxy

Learn about the Harness Forward Proxy, which allows you to establish a single connection point from your FME SDKs to Harness services by using a forward or reverse proxy mode.

Enterprise customers often operate in environments with controlled egress points, where all outgoing network traffic must pass through approved gateways. Without a proxy, each new delegate or SDK instance may require individual firewall exceptions and cross-team approvals, creating operational friction at scale.

Harness Proxy simplifies this by acting as a single, authorized egress point for all Harness-related traffic. The Harness Proxy also supports reverse proxy mode, which allows SDK traffic or other client requests to be routed from your infrastructure to Harness services through a centralized entry point.

This allows you to:

• Centralize and control inbound and outbound traffic from your infrastructure

• Maintain compliance with enterprise security policies

• Enable advanced authentication and authorization, including mTLS and JWT validation

• Support long-lived connections and encrypted payloads without additional proxy-side configuration

• Expose Harness endpoints through a centralized reverse proxy for simplified client access

The Harness Proxy is fully customer-deployed, giving you flexibility to run it in your preferred environment (Docker, Kubernetes, or custom deployments) while keeping sensitive configuration and certificates under your control.

How Harness Proxy works

The Harness Proxy runs on OpenResty, an NGINX-based platform chosen for its scalability, reliability, and flexibility. All configuration is externalized, allowing you to mount volumes into the container for:

• TLS certificates and private keys for secure tunnels

• JWKS files for JWT validation

The proxy supports forward and reverse proxy modes:

• Forward Proxy: SDKs establish opaque tunnels through the proxy to Harness FME servers. Once deployed, SDKs establish opaque tunnels through the proxy to the Harness FME servers, and all traffic is securely forwarded without inspecting payloads.

• Reverse Proxy: Clients or SDKs can connect to pre-defined URLs exposed by the proxy, which forwards requests to the appropriate Harness backend services. Reverse proxy supports mTLS and location presets, including a built-in FME preset or custom location mappings. The proxy performs local TLS termination and re-establishes a secure upstream connection, handling routing while forwarding traffic without inspecting payloads.

Security

Harness Proxy provides a layered, opt-in security model:

• TLS/mTLS: Secure tunnels are established using X.509 certificates and private keys that you provide. The proxy can request and validate client certificates, rejecting connections that fail validation.

• Authorization schemes: Each server instance can be configured with one of the following authentication methods.

  • Basic Auth: Credentials sent in the Proxy-Authorization header (requires TLS).

  • Digest Auth: Credentials hashed with server-provided parameters to avoid sending plaintext credentials.

  • Bearer Auth: JWT-based tokens signed by the user and validated with JWKS files.

This allows flexible deployment scenarios while maintaining enterprise-grade security.

Supported Transport and Protocols

The proxy supports HTTP and HTTPS tunnels, including long-lived connections such as server-sent events (SSE). Traffic remains opaque to the proxy, meaning payloads are encrypted end-to-end, and no additional proxy-side configuration is required to support new endpoints.

Setup

To get started with Docker:

1. Pull the prebuilt image: docker pull splitsoftware/fme-proxy.

2. Run the container with your desired configuration: docker run --rm --name some-fme-proxy -e HP_PROXIES=main -e HP_main_PORT=3128 -p 3128:3128 splitsoftware/fme-proxy:ubuntu.

3. Test the connectivity by running a cURL command: curl -v --proxy http://127.0.0.1:3128 -XGET https://sdk.split.io/api/version.

This starts the proxy with a single server instance. The HFP_PROXIES environment variable allows multiple server instances with different TLS/auth settings.

Configuration

Harness Proxy is fully configurable and supports multiple server instances, each with its own port, SSL, authentication, and proxy requirements. Servers are defined in the HP_PROXIES variable as a comma-separated list of identifiers (e.g., server_mtls,server_basic_auth). All server-specific configuration variables follow the pattern HP_<server_name>_<config_option>.

Global Configuration

These variables allow the proxy to scale efficiently based on deployment environment and traffic load.

Variable	Context	Description	Default	Example
HP_PROXIES	Global	Comma-separated list of server identifiers.	none / required	server_mtls,server_basic_auth
HP_WORKER_PROCESSES	Global	Number of NGINX worker processes. Ideally matches CPU threads.	4	8
HP_WORKER_CONNECTIONS	Global	Max connections per worker.	1024	2048
HP_LOG_LEVEL	Global	Maximum level of unfiltered logs.	info	debug

Server-Specific Configuration

These variables allow you to tailor each proxy instance’s ports, TLS, authentication, and routing to meet your security and traffic requirements.

Forward Proxy Server Configuration

Variable	Description	Default	Example
HP_<SERVER_NAME>_TYPE	Must be FORWARD.	FORWARD	FORWARD
HP_<SERVER_NAME>_PORT	Port on which this server listens.	Required	3128
HP_<SERVER_NAME>_SSL	Enable TLS/mTLS.	False	true
HP_<SERVER_NAME>_SSL_CERTIFICATE	Server certificate for TLS.	None / required if SSL enabled	/my_volume/pki/server.crt
HP_<SERVER_NAME>_SSL_PRIVATE_KEY	Private key for TLS certificate.	None / required if SSL enabled	/my_volume/pki/server.key
HP_<SERVER_NAME>_SSL_CLIENT_CERTIFICATE	CA certificate to verify client certificates (mTLS).	None	/my_volume/pki/client_ca.crt
HP_<SERVER_NAME>_AUTH	Authentication scheme (basic, digest, bearer).	None	bearer
HP_<SERVER_NAME>_AUTH_BASIC_PASSWD	Password file for basic auth.	None / required if auth=basic	/my_volume/passwd/basic.passwd
HP_<SERVER_NAME>_AUTH_DIGEST_PASSWD	Password file for digest auth.	None / required if auth=digest	/my_volume/passwd/digest.passwd
HP_<SERVER_NAME>_AUTH_BEARER_JWKS	JWKS file to validate bearer tokens.	None / required if auth=bearer	/my_volume/keys/keys.jwks
HP_<SERVER_NAME>_PROXY_THRU	Upstream proxy to forward traffic.	None	my-upstream-proxy:3128
HP_<SERVER_NAME>_PROXY_THRU_SSL	Use TLS to connect to upstream proxy.	Disabled	true
HP_<SERVER_NAME>_PROXY_THRU_CA_CERT	CA cert to verify upstream proxy TLS.	None / required if PROXY_THRU_SSL enabled	/my_volume/pki/ca.crt
HP_<SERVER_NAME>_RESOLVER	DNS server for upstream host resolution.	8.8.8.8	127.0.0.53
HP_<SERVER_NAME>_ALLOWED_TARGETS	Comma-separated list of allowed host:port pairs. Requests outside these hosts are rejected.	sdk.split.io:443,auth.split.io:443,streaming.split.io:443,events.split.io:443,telemetry.split.io:443	api.mycompany.com:443
HP_<SERVER_NAME>_ALLOWED_TARGET_PORTS	Comma-separated list of allowed ports. Must include all ports in ALLOWED_TARGETS.	443	443,8443

Reverse Proxy Server Configuration

Variable	Description	Default	Example
HP_<SERVER_NAME>_PORT	Port on which this server listens.	Required	3129
HP_<SERVER_NAME>_SSL	Enable TLS/mTLS.	False	true
HP_<SERVER_NAME>_SSL_CERTIFICATE	Server certificate for TLS.	None / required if SSL enabled	/my_volume/pki/server.crt
HP_<SERVER_NAME>_SSL_PRIVATE_KEY	Private key for TLS certificate.	None / required if SSL enabled	/my_volume/pki/server.key
HP_<SERVER_NAME>_SSL_CLIENT_CERTIFICATE	CA certificate to verify client certificates (mTLS).	None	/my_volume/pki/client_ca.crt
HP_<SERVER_NAME>_PROXY_THRU	Upstream proxy to forward traffic.	None	my-upstream-proxy:3128
HP_<SERVER_NAME>_PROXY_THRU_SSL	Use TLS to connect to upstream proxy.	Disabled	true
HP_<SERVER_NAME>_PROXY_THRU_CA_CERT	CA cert to verify upstream proxy TLS.	None / required if PROXY_THRU_SSL enabled	/my_volume/pki/ca.crt
HP_<SERVER_NAME>_RESOLVER	DNS server for upstream host resolution.	8.8.8.8	127.0.0.53
HP_<SERVER_NAME>_LOCATIONS	Comma-separated list of reverse proxy locations. Can reference built-in presets (PRESET:fme) or custom JSONL files.	Required	PRESET:fme,PRESET:custom_service
HP_<SERVER_NAME>_LOCATIONS_SSL_VERIFICATION_CERT	CA cert to validate SSL connections to upstream targets. Optional; TLS verification skipped if not provided.	443	/my_volume/pki/ca.crt

Reverse Proxy locations

When using a reverse proxy server, you can define locations to map incoming paths to upstream services. Locations can reference built-in Harness module presets or be fully customized.

Built-in presets allow you to configure common Harness services. For example, include PRESET:fme for FME in the HP_<SERVER_NAME>_LOCATIONS variable. This maps the following paths automatically:

Path	Target
/fme/sdk	https://sdk.split.io
/fme/events	https://events.split.io
/fme/auth	https://auth.split.io
/fme/streaming	https://streaming.split.io
/fme/telemetry	https://telemetry.split.io

These paths correspond to the URLs your Harness FME SDKs use when connecting through the reverse proxy.

Custom locations

For external services or non-standard routes, create a JSONL file with one location definition per line. Each object must include at least path and target. Additional optional parameters include buffer and read_timeout.

Property	Description	Default	Example
path	Local path clients use to reference the proxied host.	Required	/custom/api
target	URL of the upstream host where requests are forwarded.	Required	https://api.myservice.com
buffer	Whether responses should be buffered.	true	false
read_timeout	Socket read timeout in seconds (useful for streaming endpoints).	30	120

The following code snippet demonstrates an example JSONL file for FME SDKs:

{ "path": "/fme/sdk", "target": "https://sdk.split.io" }
{ "path": "/fme/events", "target": "https://events.split.io" }
{ "path": "/fme/auth", "target": "https://auth.split.io" }
{ "path": "/fme/streaming", "target": "https://streaming.split.io", "read_timeout": 120, "buffer": false }
{ "path": "/fme/telemetry", "target": "https://telemetry.split.io" }

Harness recommends following these best practices:

TLS and mTLS

• TLS is recommended when using plain-text authentication schemes (Basic or Bearer/JWT). These mechanisms exchange credentials in clear text and rely on TLS for confidentiality.

• If the proxy operates in forward mode behind a reverse proxy/ingress, enabling TLS helps avoid host header conflicts by relying on SNI instead of the Host header (common in ingress setups such as CITI).

• mTLS can be used for strong client identity verification when the proxy is deployed inside a PKI. This approach requires issuing certificates to each client but provides higher security than password or token-based methods.

Worker and Performance Settings

• Set HP_WORKER_PROCESSES to the number of vCPUs available to the container. This should be a whole integer.

• HP_WORKER_CONNECTIONS controls the number of simultaneous connections and has a large memory impact and a small CPU impact.

  • 1024 is a reliable default for most deployments.

  • Increase only when handling high concurrency, and monitor CPU and memory usage.

Outbound Restrictions (Forward Proxy)

• Use HP_<server_name>_ALLOWED_TARGETS and HP_<server_name>_ALLOWED_TARGET_PORTS to strictly limit what upstream services can be reached. This prevents tunnels from being abused for non-Harness traffic.

Authentication (Forward Proxy)

• Use HP_<server_name>_AUTH_BEARER_JWKS when enabling bearer authentication to validate JWTs using a trusted key set.

Reverse Proxy Locations

• Use HP_<server_name>_LOCATIONS to expose pre-defined Harness endpoints (e.g., using PRESET:fme) or to inject custom mappings.

• To provide custom locations, create a JSONL file following the documented location schema, mount it into the container, and reference it using the CUSTOM: prefix, for example:

  HP_<server_name>_LOCATIONS=CUSTOM:/path/to/file.jsonl

• Optionally, use HP_<server_name>_LOCATIONS_SSL_VERIFICATION_CERT to validate TLS connections to the upstream targets.

Multiple Servers

• Each configured server (forward or reverse) can have independent TLS, authentication, and port settings, allowing fine-grained traffic control.

Use FME SDKs with the Harness Proxy

Harness Proxy works with multiple FME SDKs to securely forward traffic from your infrastructure to Harness services. SDKs can connect via forward proxy tunnels or through reverse proxy endpoints defined by the proxy server. Traffic is opaque to the proxy, so encrypted payloads and long-lived connections (such as SSE) are fully supported without additional configuration.

Android SDK

Browser SDK

Java SDK

Node.js SDK

To integrate an SDK with the proxy, configure the SDK to point to the proxy host and port by setting the proxy host/port in the SDK initialization. This ensures that all SDK requests are routed through a centralized, secure, and authenticated point.