Create a Network Load Balancer

Create a Network Load Balancer (NLB) to forward TCP connections to your servers. Each backend group listens on its own port, so one NLB can carry several services, such as a mail server on one port and a message queue on another.

An NLB forwards TCP connections as they arrive. It passes encrypted traffic through to your servers as it is, and it checks each server by opening a connection to its port. For HTTP or HTTPS with an SSL certificate on the load balancer, ACL rules, or a scale group, create an Application Load Balancer instead; see Create an Application Load Balancer.

Before you begin

• You are working in Delhi or Chennai, in the project where the load balancer will live.
• You have the nodes that run your service, and you know the TCP port it listens on in each. See Create a node.
• You have chosen the port clients will connect to on the load balancer for each service: 443, or a port from 1024 to 65535, other than 8080, 10050, and 9101.
• Each server’s security group lets the load balancer in on the port your service uses. See Manage a load balancer’s network and security groups.
• For an internal load balancer, you have a VPC ready in the Active state. See Create a VPC.
• You have the create permission for Load Balancers in the project, and enough balance or infra credits to pay for the plan.

Choose the plan and the type

1. In MyAccount, under Compute, click Load Balancers. Click +, or Get Started if you have no load balancer yet.

2. On Plan, click a plan, choose Hourly Billed or a committed option under —Choose Hourly / Committed LB—, and click Next. The plans run from E2E-LB-2, with 4 vCPU and 6 GB RAM, to E2E-LB-5, with 24 vCPU and 48 GB RAM. For the plan table and the committed options, see Create an Application Load Balancer: Choose the plan.

   When you work in Delhi with an account created on or after May 15, 2026, every plan offers Request in place of Select, unless E2E has exempted your account. Click Request to ask E2E for the load balancer, and We have accepted your request, will contact you shortly. confirms it.

3. On Type, click Network Load Balancer (NLB). The type is fixed once the load balancer is created.

Fill in the details

1. Keep the suggested Name, or enter your own, up to 50 characters, using letters, digits, underscores, and hyphens.
2. Under Type, choose External (with public IP) - Distributes traffic from external sources for clients on the internet, or Internal (without public IP) - Distributes traffic within private VPC networks for clients inside your VPCs.
3. Click Next.

Set up a backend group for each port

On Target Mapping, you set up one backend group: one listening port on the load balancer with the servers behind it.

1. Keep the suggested Backend Group name, or type your own: up to 50 letters, digits, underscores, and hyphens.

2. Under Routing Policy, keep Least Connection for long-lived connections, or choose Round Robin to take the servers in turn, or Source IP Hash to keep each client on one server. See Choose a routing policy.

3. In LB Listening Port, enter the port clients connect to on the load balancer: 443, or a port from 1024 to 65535, other than 8080, 10050, and 9101. Each group needs a port of its own.

   Port 80 is also accepted, but keep NLB groups off it: about half of the connections to a group there get 503 Service Unavailable.

4. Beside Backend Servers, click Add Server for each server. Fill in Server Name, Server IP, and Server Port, the port your service listens on in that server, and click Submit. Typing a node’s name in Server Name lets you pick it and fills in its IP. If you attach a VPC on the next step, give the node’s address in that VPC; otherwise, its private IP. Each group needs at least one server, and server names within a group must differ.

LB Listening Port and Server Port are separate, so they can differ. For example, clients can connect to port 5432 on the load balancer while your servers listen on 15432.

To carry another service, add a group on Backend Mapping with Add Backend Group, each with its own listening port; see Manage backend groups and servers. Click Next.

Choose the Peak Performance Features

The options are the same as on an Application Load Balancer, and are explained in Create an Application Load Balancer: Choose the Peak Performance Features. Two of them need your attention:

• For an internal load balancer, pick at least one VPC under VPC.
• Under Choose Security Group, keep the default group or choose another. Check that it allows inbound TCP on each LB Listening Port from your clients.

The rest are optional: a reserved IP under IPV4, Bitninja, the timeouts, and Enable Encryption. Encryption, and a reserved IP as the public IP, can be chosen only at this step. Of the timeouts, an NLB uses Connection Timeout, Client Timeout, and Server Timeout.

Click Next.

Launch the load balancer

On Summary, check your choices, and click Launch Load Balancer. MyAccount takes you back to the Load Balancers list, where the new NLB reads Creating. It reads Running about four to seven minutes after you launch it. For what can stop a launch, see Create an Application Load Balancer: Launch the load balancer.

Check that the load balancer is running

Wait until the load balancer reads Running. Its Mode in the list reads TCP, and its Info tab shows the first group’s LB Listening Port. Connect a client to a listening port, where <load-balancer-ip> is the load balancer’s address and <listening-port> the port:

nc -zv -w 5 <load-balancer-ip> <listening-port>

For an external load balancer, use the address in the Public IP column. For an internal one, run the command from a machine in the same VPC, with the VPC address shown under Private IP. A line that says the connection succeeded means the load balancer accepted it on that port. When no server in the group passes its check, the NLB accepts the connection and closes it without sending any data. To see whether each server passes its check, click the load balancer and open its Backend Mapping tab.

To add services or servers later, see Manage backend groups and servers. If it reads Backend Connection Failure or a port is refused, see Troubleshooting Load Balancer.