Backend service-based network load balancer overview

An external TCP/UDP network load balancer (after this referred to as network load balancer) is a regional, pass-through load balancer. Network load balancers distribute external traffic among virtual machine (VM) instances in the same region.

You can configure a network load balancer for TCP, UDP, ESP, GRE, ICMP, and ICMPv6 traffic.

A network load balancer can receive traffic from:

  • Any client on the internet
  • Google Cloud VMs with external IPs
  • Google Cloud VMs that have internet access through Cloud NAT or instance-based NAT

Network load balancers have the following characteristics:

  • A network load balancer is a managed service.
  • Network load balancers are implemented by using Andromeda virtual networking and Google Maglev.
  • Network load balancers are not proxies.
    • Load-balanced packets are received by backend VMs with the packet's source and destination IP addresses, protocol, and, if the protocol is port-based, the source and destination ports unchanged.
    • Load-balanced connections are terminated by the backend VMs.
    • Responses from the backend VMs go directly to the clients, not back through the load balancer. The industry term for this is direct server return (DSR).

Backend service-based network load balancers have the following characteristics:

  • Managed instance group backends. Backend service-based network load balancers support using managed instance groups (MIGs) as backends. Managed instance groups automate certain aspects of backend management and provide better scalability and reliability as compared to unmanaged instance groups.
  • Support for multiple protocols. Backend service-based network load balancers can load-balance TCP, UDP, ESP, GRE, ICMP, and ICMPv6 traffic.
  • Support for IPv6 connectivity. Backend service-based network load balancers can handle both IPv4 and IPv6 traffic.
  • Fine-grained traffic distribution control. A backend service allows traffic to be distributed according to a configurable session affinity, connection tracking mode, and weighted load balancing. The backend service can also be configured to enable connection draining and designate failover backends for the load balancer. Most of these settings have default values that let you get started quickly.
  • Health checks. Backend service-based network load balancers use health checks that match the type of traffic (TCP, SSL, HTTP, HTTPS, or HTTP/2) that they are distributing.
  • Google Cloud Armor integration. Google Cloud Armor supports advanced network DDoS protection for network load balancers. For more information, see Configure advanced network DDoS protection.

Load balancing for GKE applications

If you are building applications in GKE, we recommend that you use the built-in GKE Service controller, which deploys Google Cloud load balancers on behalf of GKE users. This is the same as the standalone load balancing architecture described on this page, except that its lifecycle is fully automated and controlled by GKE.

Related GKE documentation:

Architecture

The following diagram illustrates the components of a network load balancer:

Network load balancer with a regional backend service
Network load balancer with a regional backend service

The load balancer is made up of several configuration components. A single load balancer can have the following:

  • One or more regional external IP addresses
  • One or more regional external forwarding rules
  • One regional external backend service
  • One or more backend instance groups
  • Health check associated with the backend service

Additionally, you must create firewall rules that allow your load balancing traffic and health check probes to reach the backend VMs.

IP address

A network load balancer requires at least one forwarding rule. The forwarding rule references a regional external IP address that is accessible anywhere on the internet.

  • For IPv4 traffic, the forwarding rule references a single regional external IPv4 address. Regional external IPv4 addresses come from a pool unique to each Google Cloud region. The IP address can be a reserved static address or an ephemeral address.

  • For IPv6 traffic, the forwarding rule references a /96 range from a dual-stack subnet with an external IPv6 subnet range in the VPC network. External IPv6 addresses are available only in Premium Tier. The IPv6 address range can be a reserved static address or an ephemeral address.

Use a reserved IP address for the forwarding rule if you need to keep the address associated with your project for reuse after you delete a forwarding rule or if you need multiple forwarding rules to reference the same IP address.

Network load balancers support both Standard Tier and Premium Tier for regional external IPv4 addresses. Both the IP address and the forwarding rule must use the same network tier. Regional external IPv6 addresses are only available in the Premium Tier.

Forwarding rule

A regional external forwarding rule specifies the protocol and ports on which the load balancer accepts traffic. Because network load balancers are not proxies, they pass traffic to backends on the same protocol and ports, if the packet carries port information. The forwarding rule in combination with the IP address forms the frontend of the load balancer.

The load balancer preserves the source IP addresses of incoming packets. The destination IP address for incoming packets is an IP address associated with the load balancer's forwarding rule.

Incoming traffic is matched to a forwarding rule, which is a combination of a particular IP address (either an IPv4 address or an IPv6 address range), protocol, and if the protocol is port-based, one of port(s), a range of ports, or all ports. The forwarding rule then directs traffic to the load balancer's backend service.

  • If the forwarding rule references an IPv4 address, the forwarding rule is not associated with any subnet. That is, its IP address comes from outside of any Google Cloud subnet range.

  • If the forwarding rule references a /96 IPv6 address range, the forwarding rule must be associated with a subnet, and that subnet must be (a) dual-stack and (b) have an external IPv6 subnet range (--ipv6-access-type set to EXTERNAL). The subnet that the forwarding rule references can be the same subnet used by the backend instances; however, backend instances can use a separate subnet if desired. When backend instances use a separate subnet, the following must be true:

    • The subnet used by the backend instances must be located in the same region as the forwarding rule.
    • The subnet used by the backend instances must be dual-stack; however, it can use an internal IPv6 address range (--ipv6-access-type set to INTERNAL) if desired.

A network load balancer requires at least one forwarding rule. Forwarding rules can be configured to direct traffic coming from a specific range of source IP addresses to a specific backend service (or target instance). For details, see traffic steering. You can define multiple forwarding rules for the same load balancer as described in Multiple forwarding rules.

If you want the load balancer to handle both IPv4 and IPv6 traffic, create two forwarding rules: one rule for IPv4 traffic that points to IPv4 (or dual-stack) backends, and one rule for IPv6 traffic that points only to dual-stack backends. It's possible to have an IPv4 and an IPv6 forwarding rule reference the same backend service, but the backend service must reference dual-stack backends.

Forwarding rule protocols

Network load balancers support the following protocol options for each forwarding rule: TCP, UDP, and L3_DEFAULT.

Use the TCP and UDP options to configure TCP or UDP load balancing. The L3_DEFAULT protocol option enables a network load balancer to load balance TCP, UDP, ESP, GRE, ICMP, and ICMPv6 traffic.

In addition to supporting protocols other than TCP and UDP, L3_DEFAULT makes it possible for a single forwarding rule to serve multiple protocols. For example, IPSec services typically handle some combination of ESP and UDP-based IKE and NAT-T traffic. The L3_DEFAULT option allows a single forwarding rule to be configured to process all of those protocols.

Forwarding rules using the TCP or UDP protocols can reference a backend service using either the same protocol as the forwarding rule or a backend service whose protocol is UNSPECIFIED. L3_DEFAULT forwarding rules can only reference a backend service with protocol UNSPECIFIED.

If you're using the L3_DEFAULT protocol, you must configure the forwarding rule to accept traffic on all ports. To configure all ports, either set --ports=ALL by using the Google Cloud CLI, or set allPorts to True by using the API.

The following table summarizes how to use these settings for different protocols.

Traffic to be load balanced Forwarding rule protocol Backend service protocol
TCP TCP TCP or UNSPECIFIED
L3_DEFAULT UNSPECIFIED
UDP UDP UDP or UNSPECIFIED
L3_DEFAULT UNSPECIFIED
ESP, GRE,
ICMP/ICMPv6 (echo request only)
 L3_DEFAULT UNSPECIFIED

Multiple forwarding rules

You can configure multiple regional external forwarding rules for the same network load balancer. Each forwarding rule can have a different regional external IP address, or multiple forwarding rules can have the same regional external IP address.

Configuring multiple regional external forwarding rules can be useful for these use cases:

  • You need to configure more than one external IP address for the same backend service.
  • You need to configure different protocols or non-overlapping ports or port ranges for the same external IP address.
  • You need to steer traffic from certain source IP addresses to specific load balancer backends.

Google Cloud requires that incoming packets match no more than one forwarding rule. Except for steering forwarding rules, which are discussed in the next section, two or more forwarding rules that use the same regional external IP address must have unique protocol and port combinations according to these constraints:

  • A forwarding rule configured for all ports of a protocol prevents the creation of other forwarding rules using the same protocol and IP address. Forwarding rules using TCP or UDP protocols can be configured to use all ports, or they can be configured for specific ports. For example, if you create a forwarding rule using IP address 198.51.100.1, the TCP protocol, and all ports, you cannot create any other forwarding rule using IP address 198.51.100.1and the TCP protocol. You can create two forwarding rules, both using the IP address 198.51.100.1 and the TCP protocol, if each one has unique ports or non-overlapping port ranges. For example, you can create two forwarding rules using IP address 198.51.100.1and the TCP protocol, where one forwarding rule's ports are 80,443 and the other uses the port range 81-442.
  • Only one L3_DEFAULT forwarding rule can be created per IP address. This is because the L3_DEFAULT protocol uses all ports by definition. In this context, the all ports term includes protocols without port information.

  • A single L3_DEFAULT forwarding rule can coexist with other forwarding rules that use specific protocols (TCP or UDP). The L3_DEFAULT forwarding rule can be used as a last resort when forwarding rules using the same IP address but more specific protocols exist. An L3_DEFAULT forwarding rule processes packets sent to its destination IP address if and only if the packet's destination IP address, protocol, and destination port do not match a protocol-specific forwarding rule.

    To illustrate this, consider these two scenarios. Forwarding rules in both scenarios use the same IP address 198.51.100.1.

    • Scenario 1. The first forwarding rule uses the L3_DEFAULT protocol. The second forwarding rule uses the TCP protocol and all ports. TCP packets sent to any destination port of 198.51.100.1 are processed by the second forwarding rule. Packets using different protocols are processed by the first forwarding rule.
    • Scenario 2. The first forwarding rule uses the L3_DEFAULT protocol. The second forwarding rule uses the TCP protocol and port 8080. TCP packets sent to 198.51.100.1:8080 are processed by the second forwarding rule. All other packets, including TCP packets sent to different destination ports, are processed by the first forwarding rule.

Forwarding rule selection

Google Cloud selects one or zero forwarding rules to process an incoming packet by using this elimination process, starting with the set of forwarding rule candidates which match the destination IP address of the packet:

  • Eliminate forwarding rules whose protocol doesn't match the packet's protocol, except for L3_DEFAULT forwarding rules. Forwarding rules using the L3_DEFAULT protocol are never eliminated by this step because L3_DEFAULT matches all protocols. For example, if the packet's protocol is TCP, only forwarding rules using the UDP protocol are eliminated.

  • Eliminate forwarding rules whose port doesn't match the packet's port. Forwarding rules configured for all ports are never eliminated by this step because an all ports forwarding rule matches any port.

  • If the remaining forwarding rule candidates include both L3_DEFAULT and protocol specific forwarding rules, eliminate the L3_DEFAULT forwarding rules. If the remaining forwarding rule candidates are all L3_DEFAULT forwarding rules, none are eliminated at this step.

  • At this point, either the remaining forwarding rule candidates fall into one of the following categories:

    • A single forwarding rule remains which matches the packet's destination IP address, protocol, and port, and is used to route the packet.
    • Two or more forwarding rule candidates remain which match the packet's destination IP address, protocol, and port. This means the remaining forwarding rule candidates include steering forwarding rules (discussed in the next section). Select the steering forwarding rule whose source range includes the most specific (longest prefix match) CIDR containing the packet's source IP address. If no steering forwarding rules have a source range including the packet's source IP address, select the parent forwarding rule.
    • Zero forwarding rule candidates remain and the packet is dropped.

When using multiple forwarding rules, make sure that you configure the software running on your backend VMs so that it binds to all the external IP address(es) of the load balancer's forwarding rule(s).

Traffic steering

Forwarding rules for network load balancers can be configured to direct traffic coming from a specific range of source IP addresses to a specific backend service (or target instance).

Traffic steering is useful for troubleshooting and for advanced configurations. With traffic steering, you can direct certain clients to a different set of backends, a different backend service configuration, or both. For example:

  • Traffic steering lets you create two forwarding rules which direct traffic to the same backend instance group by way of two backend services. The two backend services can be configured with different health checks, different session affinities, or different traffic distribution control policies (connection tracking, connection draining, and failover).
  • Traffic steering lets you create a forwarding rule to redirect traffic from a low-bandwidth backend service to a high-bandwidth backend service. Both backend services contain the same set of backend instances, but load-balanced with different weights using weighted load balancing.
  • Traffic steering lets you create two forwarding rules which direct traffic to different backend services, with different backend instance groups. For example, one instance group could be configured using different machine types in order to better process traffic from a certain set of source IP addresses.

Traffic steering is configured with a forwarding rule API parameter called sourceIPRanges. Forwarding rules that have at least one source IP range configured are called steering forwarding rules.

A steering forwarding rule can have a list of up to 64 source IP ranges. You can update the list of source IP ranges configured for a steering forwarding rule at any time.

Each steering forwarding rule requires that you first create a parent forwarding rule. The parent and steering forwarding rules share the same regional external IP address, IP protocol, and port information; however, the parent forwarding rule does not have any source IP address information. For example:

  • Parent forwarding rule: IP address: 198.51.100.1, IP protocol: TCP, ports: 80
  • Steering forwarding rule: IP address: 198.51.100.1, IP protocol: TCP, ports: 80, sourceIPRanges: 203.0.113.0/24

A parent forwarding rule that points to a backend service can be associated with a steering forwarding rule that points to a backend service or a target instance.

For a given parent forwarding rule, two or more steering forwarding rules can have overlapping, but not identical, source IP ranges. As an example, one steering forwarding rule can have the source IP range 203.0.113.0/24 and another steering forwarding rule for the same parent can have the source IP range 203.0.113.0.

You must delete all steering forwarding rules before you can delete the parent forwarding rule upon which they depend.

To learn how incoming packets are processed when steering forwarding rules are used, see Forwarding rule selection.

Session affinity behavior across steering changes

This section describes the conditions under which session affinity might break when the source IP ranges for steering forwarding rules are updated:

  • If an existing connection continues to match the same forwarding rule after you change the source IP ranges for a steering forwarding rule, session affinity will not break. If your change results in an existing connection matching a different forwarding rule, then:
  • Session affinity always breaks under these circumstances:
    • The newly-matched forwarding rule directs an established connection to a backend service (or target instance) which doesn't reference the previously-selected backend VM.
    • The newly-matched forwarding rule directs an established connection to a backend service which does reference the previously-selected backend VM, but the backend service is not configured to persist connections when backends are unhealthy, and the backend VM fails the backend service's health check.
  • Session affinity might break when the newly-matched forwarding rule directs an established connection to a backend service, and the backend service does reference the previously-selected VM, but the backend service's combination of session affinity and connection tracking mode results in a different connection tracking hash.

Preserving session affinity across steering changes

This section describes how to avoid breaking session affinity when the source IP ranges for steering forwarding rules are updated:

  • Steering forwarding rules pointing to backend services. If both the parent and the steering forwarding rule point to backend services, you'll need to manually make sure that the session affinity and connection tracking policy settings are identical. Google Cloud does not automatically reject configurations if they are not identical.
  • Steering forwarding rules pointing to target instances. A parent forwarding rule that points to a backend service can be associated with a steering forwarding rule that points to a target instance. In this case, the steering forwarding rule inherits session affinity and connection tracking policy settings from the parent forwarding rule.

For instructions on how to configure traffic steering, see Configure traffic steering.

Regional backend service

Each network load balancer has one regional backend service that defines the behavior of the load balancer and how traffic is distributed to its backends. The name of the backend service is the name of the network load balancer shown in the Google Cloud console.

Each backend service defines the following backend parameters:

  • Protocol. A backend service accepts traffic on the IP address and ports (if configured) specified by one or more regional external forwarding rules. The backend service passes packets to backend VMs while preserving the packet's source and destination IP addresses, protocol, and, if the protocol is port-based, the source and destination ports.

    Backend services used with network load balancers support the following protocol options: TCP, UDP, or UNSPECIFIED.

    Backend services with the UNSPECIFIED protocol can be used with any forwarding rule regardless of the forwarding rule protocol. Backend services with a specific protocol (TCP or UDP) can only be referenced by forwarding rules with the same protocol (TCP or UDP). Forwarding rules with the L3_DEFAULT protocol can only refer to backend services with the UNSPECIFIED protocol.

    See Forwarding rule protocol specification for a table with possible forwarding rule and backend service protocol combinations.

  • Traffic distribution. A backend service allows traffic to be distributed according to a configurable session affinity, connection tracking mode, and weighted load balancing. The backend service can also be configured to enable connection draining and designate failover backends for the load balancer.

  • Health check. A backend service must have an associated regional health check.

Each backend service operates in a single region and distributes traffic to the first network interface (nic0) of backend VMs. Backends must be instance groups in the same region as the backend service (and forwarding rule). The backends can be zonal unmanaged instance groups, zonal managed instance groups, or regional managed instance groups.

Backend service-based network load balancers support instance groups whose member instances use any VPC network in the same region, as long as the VPC network is in the same project as the backend service. (All VMs within a given instance group must use the same VPC network.)

If you want the load balancer to support IPv6 traffic, the backend service must reference backends that meet the requirements for handling IPv6 traffic.

Backend instance groups

A network load balancer distributes connections among backend VMs contained within managed or unmanaged instance groups. Instance groups can be regional or zonal in scope.

  • Regional managed instance groups. Use regional managed instance groups if you can deploy your software by using instance templates. Regional managed instance groups automatically distribute traffic among multiple zones, providing the best option to avoid potential issues in any given zone.

    An example deployment using a regional managed instance group is shown here. The instance group has an instance template that defines how instances should be provisioned, and each group deploys instances within three zones of the us-central1 region.

    Network load balancer with a regional managed instance group
    Network load balancer with a regional managed instance group

  • Zonal managed or unmanaged instance groups. Use zonal instance groups in different zones (in the same region) to protect against potential issues in any given zone.

    An example deployment using zonal instance groups is shown here. This load balancer provides availability across two zones.

    Network load balancer with zonal instance groups
    Network load balancer with zonal instance groups

If you want the load balancer to support IPv6 traffic, the backends must fulfil the following requirements:

  • Backends must be configured in dual-stack subnets that are in the same region as the load balancer's IPv6 forwarding rule. For the backends, you can use a subnet with the ipv6-access-type set to either EXTERNAL or INTERNAL. Using a subnet with ipv6-access-type set to INTERNAL requires you to use a separate dual-stack subnet with ipv6-access-type set to EXTERNAL for the load balancer's external forwarding rule. For instructions, see Add a dual-stack subnet.
  • Backend instance groups must be configured to be dual-stack with the --ipv6-network-tier set to PREMIUM. For instructions, see Create an instance template with IPv6 addresses.

Health checks

Network load balancers use regional health checks to determine which instances can receive new connections. Each network load balancer's backend service must be associated with a regional health check. Load balancers use health check status to determine how to route new connections to backend instances.

For more details about how Google Cloud health checks work, see How health checks work.

Network load balancers support the following types of health checks:

Health checks for other protocol traffic

Google Cloud does not offer any protocol-specific health checks beyond the ones listed here. When you use a network load balancer to load balance a protocol other than TCP, you must still run a TCP-based service on your backend VMs to provide the required health check information.

For example, if you are load balancing UDP traffic, client requests are load balanced by using the UDP protocol, and you must run a TCP service to provide information to Google Cloud health check probers. To achieve this, you can run a simple HTTP server on each backend VM that returns an HTTP 200 response to health check probers. You should use your own logic running on the backend VM to ensure that the HTTP server returns 200 only if the UDP service is properly configured and running.

Firewall rules

Because network load balancers are passthrough load balancers, you control access to the load balancer's backends using Google Cloud firewall rules. You must create ingress allow firewall rules or an ingress allow hierarchical firewall policy to permit health checks and the traffic that you're load balancing.

Forwarding rules and ingress allow firewall rules or hierarchical firewall policies work together in the following way: a forwarding rule specifies the protocol and, if defined, port requirements that a packet must meet to be forwarded to a backend VM. Ingress allow firewall rules control whether the forwarded packets are delivered to the VM or dropped. All VPC networks have an implied deny ingress firewall rule that blocks incoming packets from any source. The Google Cloud default VPC network includes a limited set of pre-populated ingress allow firewall rules.

  • To accept traffic from any IP address on the internet, you must create an ingress allow firewall rule with the 0.0.0.0/0 source range. To only allow traffic from certain IP address ranges, use more restrictive source ranges.

  • As a security best practice, your ingress allow firewall rules should only permit the IP protocols and ports that you need. Restricting the protocol (and, if possible, port) configuration is especially important when using forwarding rules whose protocol is set to L3_DEFAULT. L3_DEFAULT forwarding rules forward packets for all supported IP protocols (on all ports if the protocol and packet have port information).

  • Network load balancers use Google Cloud health checks. Therefore, you must always allow traffic from the health check IP address ranges. These ingress allow firewall rules can be made specific to the protocol and ports of the load balancer's health check.

IP addresses for request and return packets

When a backend VM receives a load-balanced packet from a client, the packet's source and destination are:

  • Source: the external IP address associated with a Google Cloud VM or internet-routable IP address of a system connecting to the load balancer.
  • Destination: the IP address of the load balancer's forwarding rule.

Because the load balancer is a pass-through load balancer (not a proxy), packets arrive bearing the destination IP address of the load balancer's forwarding rule. Software running on backend VMs should be configured to do the following:

  • Listen on (bind to) the load balancer's forwarding rule IP address or any IP address (0.0.0.0 or ::)
  • If the load balancer forwarding rule's protocol supports ports: Listen on (bind to) a port that's included in the load balancer's forwarding rule

Return packets are sent directly from the load balancer's backend VMs to the client. The return packet's source and destination IP addresses depend on the protocol:

  • TCP is connection-oriented so backend VMs must reply with packets whose source IP addresses match the forwarding rule's IP address so that the client can associate the response packets with the appropriate TCP connection.
  • UDP, ESP, GRE, and ICMP are connectionless. Backend VMs can send response packets whose source IP addresses either match the forwarding rule's IP address or match any assigned external IP address for the VM. Practically speaking, most clients expect the response to come from the same IP address to which they sent packets.

The following table summarizes sources and destinations for response packets:

Traffic type Source Destination
TCP The IP address of the load balancer's forwarding rule The requesting packet's source
UDP, ESP, GRE, ICMP For most use cases, the IP address of the load balancer's forwarding rule † The requesting packet's source.

† When a VM has an external IP address or when you are using Cloud NAT, it is also possible to set the response packet's source IP address to the VM NIC's primary internal IPv4 address. Google Cloud or Cloud NAT changes the response packet's source IP address to either the NIC's external IPv4 address or a Cloud NAT external IPv4 address in order to send the response packet to the client's external IP address. Not using the forwarding rule's IP address as a source is an advanced scenario because the client receives a response packet from an external IP address that does not match the IP address to which it sent a request packet.

Return path

Network load balancers use special routes outside of your VPC network to direct incoming requests and health check probes to each backend VM.

The load balancer preserves the source IP addresses of packets. Responses from the backend VMs go directly to the clients, not back through the load balancer. The industry term for this is direct server return.

Shared VPC architecture

Except for the IP address, all of the components of a network load balancer must exist in the same project. The following table summarizes Shared VPC components for network load balancers:

IP address Forwarding rule Backend components
A regional external IP address must be defined in either the same project as the load balancer or the Shared VPC host project. A regional external forwarding rule must be defined in the same project as the instances in the backend service.

The regional backend service must be defined in the same project and same region where the instances in the backend instance group exist.

Health checks associated with the backend service must be defined in the same project and the same region as the backend service.

Traffic distribution

The way that network load balancers distribute new connections depends on whether you have configured failover:

  • If you haven't configured failover, a network load balancer distributes new connections to its healthy backend VMs if at least one backend VM is healthy. When all backend VMs are unhealthy, the load balancer distributes new connections among all backends as a last resort. In this situation, the load balancer routes each new connection to an unhealthy backend VM.
  • If you have configured failover, a network load balancer distributes new connections among healthy backend VMs in its active pool, according to a failover policy that you configure. When all backend VMs are unhealthy, you can choose from one of the following behaviors:
    • (Default) The load balancer distributes traffic to only the primary VMs. This is done as a last resort. The backup VMs are excluded from this last-resort distribution of connections.
    • The load balancer drops traffic.

For details about how connections are distributed, see the next section Backend selection and connection tracking.

For details about how failover works, see the Failover section.

Backend selection and connection tracking

Network load balancers use configurable backend selection and connection tracking algorithms to determine how traffic is distributed to backend VMs.

Network load balancers use the following algorithm to distribute packets among backend VMs (in its active pool, if you have configured failover):

  1. If the load balancer has an entry in its connection tracking table matching the characteristics of an incoming packet, the packet is sent to the backend specified by the connection tracking table entry. The packet is considered to be part of a previously established connection, so the packet is sent to the backend VM that the load balancer previously determined and recorded in its connection tracking table.

  2. If the load balancer receives a packet for which it has no connection tracking entry, the load balancer does the following:

    1. The load balancer selects a backend. The load balancer calculates a hash based on the configured session affinity. It uses this hash to select a backend from among the ones that are currently healthy (unless all backends are unhealthy, in which case all backends are considered as long as the failover policy hasn't been configured to drop traffic in this situation). The default session affinity, NONE, uses the following hash algorithms:

      • For TCP and unfragmented UDP packets, a 5-tuple hash of the packet's source IP address, source port, destination IP address, destination port, and the protocol.
      • For fragmented UDP packets and all other protocols, a 3-tuple hash of the packet's source IP address, destination IP address, and the protocol.

      Backend selection can be customized by using a hash algorithm that uses fewer pieces of information. For all the supported options, see session affinity options.

      In addition, note the following:

      If you enable weighted load balancing, the hash based backend selection becomes weighted based on the weights reported by backend instances. For example:

      • Consider a backend service configured with session affinity set to NONE and a forwarding rule with protocol UDP. If there are two healthy backend instances with weights 1 and 4, then the backends will get 20% and 80% of the UDP packets respectively.
      • Consider a backend service that is configured with 3-tuple session affinity and connection tracking. The session affinity is CLIENT_IP_PROTO and connection tracking mode is PER_SESSION. If there are three healthy backend instances with weights 0, 2 and 6, then the backends will get traffic for 0%, 25%, and 75% of the new source IP addresses (the source IP addresses for which there are no existing connection tracking table entries) respectively. Traffic for existing connections will go to the previously assigned backends.

    2. The load balancer adds an entry to its connection tracking table. This entry records the selected backend for the packet's connection so that all future packets from this connection are sent to the same backend. Whether connection tracking is used depends on the protocol:

      • TCP packets. Connection tracking is always enabled, and cannot be turned off. By default, connection tracking is 5-tuple, but it can be configured to be less than 5-tuple. When it is 5-tuple, TCP SYN packets are treated differently. Unlike non-SYN packets, they discard any matching connection tracking entry and always select a new backend.

        The default 5-tuple connection tracking is used when:

        • tracking mode is PER_CONNECTION (all session affinities), or,
        • tracking mode is PER_SESSION and the session affinity is NONE, or,
        • tracking mode is PER_SESSION and the session affinity is CLIENT_IP_PORT_PROTO.

      • UDP, ESP, and GRE packets. Connection tracking is enabled only if session affinity is set to something other than NONE.

      • ICMP and ICMPv6 packets. Connection tracking cannot be used.

      For additional details about when connection tracking is enabled, and what tracking method is used when connection tracking is enabled, see connection tracking mode.

      In addition, note the following:

      • An entry in the connection tracking table expires 60 seconds after the load balancer processes the last packet that matched the entry. This 60-second idle timeout value is not configurable.
      • Depending on the protocol, the load balancer might remove connection tracking table entries when backends become unhealthy. For details and how to customize this behavior, see Connection persistence on unhealthy backends.

Session affinity options

Session affinity controls the distribution of new connections from clients to the load balancer's backend VMs. Session affinity is specified for the entire regional external backend service, not on a per backend instance group basis.

Network load balancers support the following session affinity options:

  • None (NONE). 5-tuple hash of source IP address, source port, protocol, destination IP address, and destination port
  • Client IP, Destination IP (CLIENT_IP). 2-tuple hash of source IP address and destination IP address
  • Client IP, Destination IP, Protocol (CLIENT_IP_PROTO). 3-tuple hash of source IP address, destination IP address, and protocol
  • Client IP, Client Port, Destination IP, Destination Port, Protocol (CLIENT_IP_PORT_PROTO). 5-tuple hash of source IP address, source port, protocol, destination IP address, and destination port

To learn how these session affinity options affect the backend selection and connection tracking methods, see this table.

Connection tracking mode

Whether connection tracking is enabled depends only on the protocol of the load-balanced traffic and the session affinity settings. Tracking mode specifies the connection tracking algorithm to be used when connection tracking is enabled. There are two tracking modes: PER_CONNECTION (default) and PER_SESSION.

  • PER_CONNECTION (default). This is the default tracking mode. With this connection tracking mode, TCP traffic is always tracked per 5-tuple, regardless of the session affinity setting. For UDP, ESP, and GRE traffic, connection tracking is enabled when the selected session affinity is not NONE. UDP, ESP, and GRE packets are tracked using the tracking methods described in this table.

  • PER_SESSION. If session affinity is CLIENT_IP or CLIENT_IP_PROTO, configuring this mode results in 2-tuple and 3-tuple connection tracking, respectively, for all protocols (except ICMP and ICMPv6, which are not connection-trackable). For other session affinity settings, PER_SESSION mode behaves identically to PER_CONNECTION mode.

To learn how these tracking modes work with different session affinity settings for each protocol, see the following table.

Backend selection Connection tracking mode
Session affinity setting Hash method for backend selection PER_CONNECTION (default) PER_SESSION
Default: No session affinity

(NONE)

TCP and unfragmented UDP: 5-tuple hash

Fragmented UDP and all other protocols: 3-tuple hash
  • TCP: 5-tuple connection tracking
  • All other protocols: connection tracking off
  • TCP: 5-tuple connection tracking
  • All other protocols: connection tracking off
Client IP, Destination IP

(CLIENT_IP)

 All protocols: 2-tuple hash
  • TCP and unfragmented UDP: 5-tuple connection tracking
  • Fragmented UDP, ESP, and GRE: 3-tuple connection tracking
  • All other protocols: connection tracking off
  • TCP, UDP, ESP, GRE: 2-tuple connection tracking
  • All other protocols: connection tracking off
Client IP, Destination IP, Protocol

(CLIENT_IP_PROTO)

 All protocols: 3-tuple hash
  • TCP and unfragmented UDP: 5-tuple connection tracking
  • Fragmented UDP, ESP, and GRE: 3-tuple connection tracking
  • All other protocols: connection tracking off
  • TCP, UDP, ESP, GRE: 3-tuple connection tracking
  • All other protocols: connection tracking off
Client IP, Client Port, Destination IP, Destination Port, Protocol

(CLIENT_IP_PORT_PROTO)

TCP and unfragmented UDP: 5-tuple hash

Fragmented UDP and all other protocols: 3-tuple hash
  • TCP and unfragmented UDP: 5-tuple connection tracking
  • Fragmented UDP, ESP, and GRE: 3-tuple connection tracking
  • All other protocols: connection tracking off
  • TCP and unfragmented UDP: 5-tuple connection tracking
  • Fragmented UDP, ESP, and GRE: 3-tuple connection tracking
  • All other protocols: connection tracking off

To learn how to change the connection tracking mode, see Configure a connection tracking policy.

Connection persistence on unhealthy backends

The connection persistence settings control whether an existing connection persists on a selected backend after that backend becomes unhealthy (as long as the backend remains in the load balancer's configured backend instance group).

The behavior described in this section does not apply to cases where you remove a backend VM from its instance group, or remove the instance group from the backend service. In such cases, established connections only persist as described in connection draining.

The following connection persistence options are available:

  • DEFAULT_FOR_PROTOCOL (default)
  • NEVER_PERSIST
  • ALWAYS_PERSIST

The following table summarizes connection persistence options and how connections persist for different protocols, session affinity options, and tracking modes.

Connection persistence on unhealthy backends option Connection tracking mode
PER_CONNECTION PER_SESSION
DEFAULT_FOR_PROTOCOL

TCP: connections persist on unhealthy backends (all session affinities)

All other protocols: connections never persist on unhealthy backends

TCP: connections persist on unhealthy backends if session affinity is NONE or CLIENT_IP_PORT_PROTO

All other protocols: connections never persist on unhealthy backends
NEVER_PERSIST All protocols: connections never persist on unhealthy backends
ALWAYS_PERSIST

TCP: connections persist on unhealthy backends (all session affinities)

ESP, GRE, UDP: connections persist on unhealthy backends if session affinity is not NONE

ICMP, ICMPv6: not applicable because they are not connection-trackable

This option should only be used for advanced use cases.

 Configuration not possible

TCP connection persistence behavior on unhealthy backends

Whenever a TCP connection with 5-tuple tracking persists on an unhealthy backend:

  • If the unhealthy backend continues to respond to packets, the connection continues until it is reset or closed (by either the unhealthy backend or the client).
  • If the unhealthy backend sends a TCP reset (RST) packet or does not respond to packets, then the client might retry with a new connection, letting the load balancer select a different, healthy backend. TCP SYN packets always select a new, healthy backend.

To learn how to change connection persistence behavior, see Configure a connection tracking policy.

Weighted load balancing

You can configure a network load balancer to distribute traffic across the load balancer's backend instances based on the weights reported by an HTTP health check using weighted load balancing.

Weighted load balancing requires that you configure both of the following:

  • You must set the locality load balancer policy (localityLbPolicy) of the backend service to WEIGHTED_MAGLEV.
  • You must configure the backend service with an HTTP health check. The HTTP health check responses must contain a custom HTTP response header field X-Load-Balancing-Endpoint-Weight to specify the weights with values from 0 to 1000 for each backend instance.

If you use the same instance group as a backend for multiple backend service-based network load balancers using weighted load balancing, it's recommended to use a unique request-path for each health check of the backend service. This allows the endpoint instance to assign different weights to different backend services. For more information, see Success criteria for HTTP, HTTPS, and HTTP/2 health checks.

For selecting a backend for a new connection, backends are assigned a strict priority order based on their weight and health status as shown in the following table:

Weight Healthy Backend Selection Priority
Weight greater than zero Yes 4
Weight greater than zero No 3
Weight equals zero Yes 2
Weight equals zero No 1

Only the highest priority backends are eligible for backend selection. If all eligible backends have zero weight, the load balancer distributes new connections among all eligible backends, treating them with equal weight. For examples of weighted load balancing, see Backend selection and connection tracking.

Weighted load balancing can be used in the following scenarios:

  • If some connections process more data than others, or some connections live longer than others, the backend load distribution might get uneven. By signaling a lower per-instance weight, an instance with high load can reduce its share of new connections, while it keeps servicing existing connections.

  • If a backend is overloaded and assigning more connections might break existing connections, such backends assign zero weight to itself. By signaling zero weight, a backend instance stops servicing new connections, but continues to service existing ones.

  • If a backend is draining existing connections before maintenance, it assigns zero weight to itself. By signaling zero weight, the backend instance stops servicing new connections, but continues to service existing ones.

For more information, see Configure weighted load balancing.

Connection draining

Connection draining is a process applied to established connections when:

  • a backend VM is removed from an instance group, or,
  • when a managed instance group removes a backend VM (by replacement, abandonment, when rolling upgrades, or scaling down), or,
  • when an instance group is removed from a backend service.

By default, connection draining is disabled. When disabled, established connections are terminated as quickly as possible. When connection draining is enabled, established connections are allowed to persist for a configurable timeout, after which the backend VM instance is terminated.

For more details about how connection draining is triggered and how to enable connection draining, see Enabling connection draining.

UDP fragmentation

Backend service-based network load balancers can process both fragmented and unfragmented UDP packets. If your application uses fragmented UDP packets, keep the following in mind:

  • UDP packets might become fragmented before reaching a Google Cloud VPC network.
  • Google Cloud VPC networks forward UDP fragments as they arrive (without waiting for all fragments to arrive).
  • Non-Google Cloud networks and on-premises network equipment might forward UDP fragments as they arrive, delay fragmented UDP packets until all fragments have arrived, or discard fragmented UDP packets. For details, see the documentation for the network provider or network equipment.

If you expect fragmented UDP packets and need to route them to the same backends, use the following forwarding rule and backend service configuration parameters:

  • Forwarding rule configuration: Use only one UDP or L3_DEFAULT forwarding rule per load-balanced IP address, and configure the forwarding rule to accept traffic on all ports. This ensures that all fragments arrive at the same forwarding rule. Even though the fragmented packets (other than the first fragment) lack a destination port, configuring the forwarding rule to process traffic for all ports also configures it to receive UDP fragments that have no port information. To configure all ports, either use the Google Cloud CLI to set --ports=ALL or use the API to set allPorts to True.

  • Backend service configuration: Set the backend service's session affinity to CLIENT_IP (2-tuple hash) or CLIENT_IP_PROTO (3-tuple hash) so that the same backend is selected for UDP packets that include port information and UDP fragments (other than the first fragment) that lack port information. Set the backend service's connection tracking mode to PER_SESSION so that the connection tracking table entries are built by using the same 2-tuple or 3-tuple hashes.

Using target instances as backends

If you're using target instances as backends for the network load balancer and you expect fragmented UDP packets, use only one UDP or L3_DEFAULT forwarding rule per IP address, and configure the forwarding rule to accept traffic on all ports. This ensures that all fragments arrive at the same forwarding rule even if they don't have the same destination port. To configure all ports, either set --ports=ALL using gcloud, or set allPorts to True using the API.

Failover

You can configure a network load balancer to distribute connections among virtual machine (VM) instances in primary backend instance groups, and then switch, if needed, to using failover backend instance groups. Failover provides yet another method of increasing availability, while also giving you greater control over how to manage your workload when your primary backend VMs aren't healthy.

By default, when you add a backend to a network load balancer's backend service, that backend is a primary backend. You can designate a backend to be a failover backend when you add it to the load balancer's backend service, or by editing the backend service later.

For more details about how failover works, see Failover overview for network load balancers.

Limitations

  • Network endpoint groups (NEGs) are not supported as backends for network load balancers.

  • You cannot use the Google Cloud console to do the following tasks:

    • Create or modify a network load balancer whose forwarding rule uses the L3_DEFAULT protocol.
    • Create or modify a network load balancer whose backend service protocol is set to UNSPECIFIED.
    • Create or modify a network load balancer that configures a connection tracking policy.
    • Create or modify source IP-based traffic steering for a forwarding rule.

    Use either the Google Cloud CLI or the REST API instead.

  • Network load balancers do not support VPC Network Peering.

What's next