About Private Service Connect backends

You can access Google APIs and published services by creating a Private Service Connect endpoint (based on a forwarding rule) or a Private Service Connect backend (based on a load balancer). This guide focuses on Private Service Connect backends.

Private Service Connect backends use a load balancer configured with Private Service Connect network endpoint group (NEG) backends. This configuration was previously referred to as a Private Service Connect endpoint with consumer HTTP(S) service controls.

Accessing APIs and services through a consumer-managed load balancer provides several benefits. Load balancers can act as a centralized policy enforcement point where security policies (such as Google Cloud Armor policies and SSL policies) or routing policies (such as Google Cloud URL maps) are enforced. They provide centralized metrics and logging that a published service might not provide, and they allow consumers to control their own routing and failover.

Figure 1 shows a load balancer with a Private Service Connect NEG connecting to a published service. Client traffic goes to a load balancer that processes the traffic and then routes it to a Private Service Connect backend that maps to a published service that runs in a different VPC network.

**Figure 1.** Using a global external Application Load Balancer lets service consumers with internet access send traffic to services in the service producer's VPC network (click to enlarge).

Deployment overview

To access APIs and services through Private Service Connect backends, do the following:

Identify the API or service that you want to connect to.

For Google APIs: Select a regional service endpoint.

For published services: Ask the service producer for the service attachment URI.
Deploy a load balancer to send traffic to your published service. Choose a load balancer that fits your requirements, including whether you have internet clients, internal clients, or require regional isolation. You can also reuse an existing load balancer.
Deploy the Private Service Connect NEGs and add them to your load balancer backend service. Create Private Service Connect NEGs that reference your published service. Then add the NEGs to the load balancer's backend service so that the load balancer can send them traffic.

Supported load balancers and targets

You can use a backend to access a published service or a supported Google API.

See the load balancing documentation for more information about the load balancer that you want to add a Private Service Connect backend to.

For information about global external Application Load Balancers and regional external Application Load Balancers, see External Application Load Balancer overview.
For information about internal Application Load Balancers and Cross-region internal Application Load Balancers, see Internal Application Load Balancer overview.
For information about regional internal proxy Network Load Balancers, see Regional internal proxy Network Load Balancer overview.
For information about regional external proxy Network Load Balancers, see Regional external proxy Network Load Balancer overview.
For information about global external proxy Network Load Balancers, see External proxy Network Load Balancer overview.

Published service targets

A Private Service Connect backend for published services requires two load balancers—a consumer load balancer and a producer load balancer.

Consumer configuration

This table describes the consumer load balancers that are supported by Private Service Connect backends for published services, including which backend service protocols can be used with each consumer load balancer. The consumer load balancers can access published services that are hosted on supported producer load balancers.

Consumer load balancer	Protocols	IP version
Cross-region internal Application Load Balancer	HTTP HTTPS HTTP2	IPv4
Cross-region internal proxy Network Load Balancer	TCP	IPv4
Global external Application Load Balancer Note: Classic Application Load Balancer is not supported.	HTTP HTTPS HTTP2	IPv4
Global external proxy Network Load Balancer To associate this load balancer with a Private Service Connect NEG, use the Google Cloud CLI or send an API request. Note: Classic proxy Network Load Balancer is not supported.	TCP/SSL	IPv4
Regional external Application Load Balancer	HTTP HTTPS HTTP2	IPv4
Regional external proxy Network Load Balancer	TCP	IPv4
Regional internal Application Load Balancer	HTTP HTTPS HTTP2	IPv4
Regional internal proxy Network Load Balancer	TCP	IPv4

Producer configuration

This table describes the configuration for producer load balancers that are supported by Private Service Connect backends for published services.

Producer type	Producer configuration (published service)
Producer type	Supported producer backends	Forwarding rule protocols	Forwarding rule ports	PROXY protocol	IP version	Private Service Connect health support
Cross-region internal Application Load Balancer	GCE_VM_IP_PORT zonal NEGs Hybrid NEGs Serverless NEGs Private Service Connect NEGs Instance groups	TCP HTTP HTTPS HTTP/2 gRPC	Supports one, multiple, or all ports		IPv4
Internal passthrough Network Load Balancer	GCE_VM_IP zonal NEGs Instance groups	TCP	See Producer port configuration		IPv4
Regional internal Application Load Balancer	GCE_VM_IP_PORT zonal NEGs Hybrid NEGs Serverless NEGs Private Service Connect NEGs Instance groups	HTTP HTTPS HTTP/2	Supports a single port		IPv4
Regional internal proxy Network Load Balancer	GCE_VM_IP_PORT zonal NEGs Hybrid NEGs Private Service Connect NEGs Instance groups	TCP	Supports a single port		IPv4
Secure Web Proxy	Not applicable	Not applicable	Not applicable		IPv4

For an example backend configuration that uses a global external Application Load Balancer, see Access published services through backends.

Regional Google API targets

This table describes which load balancers can use a Private Service Connect backend to access regional Google APIs.

For an example configuration that uses an internal Application Load Balancer, see Access Google APIs through backends.

Configuration	Details
Consumer configuration (Private Service Connect backend)
Supported consumer load balancers	Internal Application Load Balancer Protocols: HTTPS Regional external Application Load Balancer Protocols: HTTPS
IP version	IPv4
Producer
Supported services	Supported regional Google APIs

Global Google API targets

This table describes which load balancers can use a Private Service Connect backend to access a global Google API.

Configuration	Details
Consumer configuration (Private Service Connect backend)
Supported consumer load balancers	Global external Application Load Balancer Note: Classic Application Load Balancer is not supported. Cross-region internal Application Load Balancer
IP version	IPv4
Producer
Supported services	Bigtable: `bigtable.googleapis.com` and `bigtableadmin.googleapis.com` Cloud Logging: `logging.googleapis.com` Spanner: `spanner.googleapis.com` Cloud Storage: `storage.googleapis.com` Pub/Sub: `pubsub.googleapis.com`

Connection statuses

Private Service Connect endpoints, backends, and service attachments have a connection status that describes the state of their connection. The consumer and producer resources that form the two sides of a connection always have the same status. You can view connection statuses when you view endpoint details, describe a backend, or view details for a published service.

The following table describes the possible statuses.

Connection status	Description
Accepted	The Private Service Connect connection is established. The two VPC networks have connectivity, and the connection is functioning normally.
Pending	The Private Service Connect connection is not established, and network traffic can't travel between the two networks. A connection might have this status for the following reasons: The service attachment requires explicit approval, and the consumer is not in the consumer accept list. The number of connections exceeds the service attachment's connection limit. The number of propagated connections that are associated with an endpoint exceeds the service attachment's propagated connection limit. Connections that are blocked for these reasons remain in the pending state indefinitely until the underlying issue is resolved.
Rejected	The Private Service Connect connection is not established. Network traffic can't travel between the two networks. A connection might have this status for the following reasons: A producer organization policy rejected the connection. A consumer reject list rejected the connection.
Needs attention	There is an issue on the producer side of the connection. Some traffic might be able to flow between the two networks, but some connections might not be functional. For example, the producer's NAT subnet might be exhausted and unable to allocate IP addresses for new connections.
Closed	The service attachment was deleted, and the Private Service Connect connection is closed. Network traffic can't travel between the two networks. A closed connection is a terminal state. To restore the connection, you must recreate both the service attachment and the endpoint or backend.

Specifications

All Private Service Connect backends have the following specifications:

Only the supported load balancers can use Private Service Connect NEGs as backends.
Private Service Connect NEGs cannot be mixed with other NEG types in the same backend service. However, self-hosted applications and managed services can both be backends of the same load balancer as long as they are part of separate backend services.
Backend services with Private Service Connect NEGs don't support health checks. Health check resources are not configured with backend services used for Private Service Connect.
Backend services with Private Service Connect NEGs don't support session affinity.
If a Private Service Connect NEG references a service attachment, the service attachment must be in a different VPC network from the NEG and the load balancer.
Private Service Connect NEGs can't reference service attachments that are configured for port mapping services.

Private Service Connect backends that are used in global backend services have additional specifications:

Multiple Private Service Connect NEGs can be in the same backend service as long as they are from different regions. You can't add multiple Private Service Connect NEGs from the same region to the same backend service.
You can take advantage of automatic cross-region failover by associating multiple Private Service Connect NEGs with the same backend service. For more information, see the following section.

Automatic cross-region failover

Accessing published services using Private Service Connect backends that are based on global or cross-regional load balancers lets you take advantage of automatic cross-region failover.

With automatic failover, if a service instance in one region becomes unhealthy, your consumer load balancer stops routing traffic to the unhealthy instance and instead routes traffic to a healthy service instance in an alternate region.

To support automatic failover, both the service producer and the service consumer must configure their resources for a multi-region deployment, as described in this section. For information about additional producer requirements for failover with Private Service Connect health, see Private Service Connect health specifications.

Producer configuration:

Deploy the service in each region. Each regional instance of the service must be configured on a regional load balancer that supports access by a backend.
Create a service attachment to publish each regional instance of the service.

Consumer configuration:

Create a Private Service Connect backend to access published services. The backend must be based on a load balancer that supports cross-region failover and includes the following configuration:
- A Private Service Connect NEG in each region that points to that region's service attachment
- A global backend service that contains the NEG backends

A global external Application Load Balancer with multiple Private Service Connect NEGs connects to a service that is published in multiple regions. This multi-region deployment lets the consumer load balancer fail over when a service instance becomes unhealthy, routing traffic to a healthy service instance in an alternate region (click to enlarge).

Automatic failover can be triggered in two ways:

Failover with outlier detection: The load balancer's standard failover mechanism, which is enabled by default in multi-region deployments. Traffic is directed away from Private Service Connect NEGs that receive a high rate of errors from the published service.
Enhanced failover with Private Service Connect health: Service producers can configure Private Service Connect health to provide more detailed health signals for their services.

Failover through outlier detection

When multiple Private Service Connect NEGs are configured in a global backend service, outlier detection is automatically enabled on the backend service.

When outlier detection identifies failures in responses sent by a published service, such as 5xx response codes, the consumer load balancer fails over, temporarily redirecting traffic to a healthy service instance in an alternate region.

You can replace the default outlier detection policy by applying your own outlier detection configuration to the backend service, or you can disable the feature by configuring a single Private Service Connect NEG in the backend service and routing 100% of your traffic to this NEG.

Enhanced failover through Private Service Connect health

With Private Service Connect health, a consumer load balancer can fail over based on a direct health signal that is configured by the service producer.

The producer defines conditions that create a single composite health state for each regional published service instance. The composite health state is based on the health of the service's backends, such as VM instances or network endpoints. For example, a producer can specify that their service is considered healthy only when a certain percentage of its backend instances are healthy.

For supported load balancers in multi-region deployments, no additional configuration is required from consumers to use health signals from Private Service Connect health.

For information about how service producers can configure Private Service Connect health, see About Private Service Connect health.

Pricing

For pricing information, see the following sections of the VPC pricing page: