Services or capabilities described in Amazon Web Services documentation might vary by Region. To see the differences applicable to the China Regions, see Getting Started with Amazon Web Services in China (PDF).

Troubleshoot your Network Load Balancer

The following information can help you troubleshoot issues with your Network Load Balancer.

A registered target is not in service

If a target is taking longer than expected to enter the InService state, it might be failing health checks. Your target is not in service until it passes one health check. For more information, see Health checks for your target groups.

Verify that your instance is failing health checks and then check for the following:

Requests are not routed to targets

Check for the following:

Targets receive more health check requests than expected

Health checks for a Network Load Balancer are distributed and use a consensus mechanism to determine target health. Therefore, targets receive more than the number of health checks configured through the HealthCheckIntervalSeconds setting.

Targets receive fewer health check requests than expected

Check whether net.ipv4.tcp_tw_recycle is enabled. This setting is known to cause issues with load balancers. The net.ipv4.tcp_tw_reuse setting is considered a safer alternative.

Unhealthy targets receive requests from the load balancer

This occurs when all registered targets are unhealthy. If there is at least one healthy registered target, your Network Load Balancer routes requests only to its healthy registered targets.

When there are only unhealthy registered targets, the Network Load Balancer routes requests to all the registered targets, known as fail-open mode. The Network Load Balancer does this instead of removing all the IP addresses from DNS when all the targets are unhealthy and respective Availability Zones do not have healthy target to send request to.

Target fails HTTP or HTTPS health checks due to host header mismatch

The HTTP host header in the health check request contains the IP address of the load balancer node and the listener port, not the IP address of the target and the health check port. If you are mapping incoming requests by host header, you must ensure that health checks match any HTTP host header. Another option is to add a separate HTTP service on a different port and configure the target group to use that port for health checks instead. Alternatively, consider using TCP health checks.

Unable to associate a security group with a load balancer

If the Network Load Balancer was created without security groups, it can't support security groups after creation. You can only associate a security group to a load balancer during creation, or to an existing load balancer that was originally created with security groups.

Unable to remove all security groups

If the Network Load Balancer was created with security groups, there must be at least one security group associated with it at all times. You cannot remove all security groups from the load balancer at the same time.

Increase in TCP_ELB_Reset_Count metric

For each TCP request that a client makes through a Network Load Balancer, the state of that connection is tracked. If no data is sent through the connection by either the client or the target for longer than the idle timeout, the connection is closed. If a client or a target sends data after the idle timeout period elapses, it receives a TCP RST packet to indicate that the connection is no longer valid. Additionally, if a target becomes unhealthy, the load balancer sends a TCP RST for packets received on the client connections associated with the target, unless the unhealthy target triggers the load balancer to fail open.

If you see a spike in the TCP_ELB_Reset_Count metric just before or just as the UnhealthyHostCount metric increases, it is likely that the TCP RST packets were sent because the target was starting to fail but hadn't been marked unhealthy. If you see persistent increases in TCP_ELB_Reset_Count without targets being marked unhealthy, you can check the VPC flow logs for clients sending data on expired flows.

Connections time out for requests from a target to its load balancer

Check whether client IP preservation is enabled on your target group. NAT loopback, also known as hairpinning, is not supported when client IP preservation is enabled. If an instance is a client of a load balancer that it's registered with, and it has client IP preservation enabled, the connection succeeds only if the request is routed to a different instance. If the request is routed to the same instance it was sent from, the connection times out because the source and destination IP addresses are the same.

If an instance must send requests to a load balancer that it's registered with, do one of the following:

Performance decreases when moving targets to a Network Load Balancer

Both Classic Load Balancers and Application Load Balancers use connection multiplexing, but Network Load Balancers do not. Therefore, your targets can receive more TCP connections behind a Network Load Balancer. Be sure that your targets are prepared to handle the volume of connection requests they might receive.

Port allocation errors connecting through Amazon PrivateLink

If your Network Load Balancer is associated with a VPC endpoint service, it supports 55,000 simultaneous connections or about 55,000 connections per minute to each unique target (IP address and port). If you exceed these connections, there is an increased chance of port allocation errors. Port allocation errors can be tracked using the PortAllocationErrorCount metric. To fix port allocation errors, add more targets to the target group. For more information, see CloudWatch metrics for your Network Load Balancer.

Intermittent connection failure when client IP preservation is enabled

When client IP preservation is enabled, you might encounter TCP/IP connection limitations related to observed socket reuse on the targets. These connection limitations can occur when a client, or a NAT device in front of the client, uses the same source IP address and source port when connecting to multiple load balancer nodes simultaneously. If the load balancer routes these connections to the same target, the connections appear to the target as if they come from the same source socket, which results in connection errors. If this happens, clients can retry (if the connection fails) or reconnect (if the connection is interrupted). You can reduce this type of connection error by increasing the number of source ephemeral ports or by increasing the number of targets for the load balancer. You can prevent this type of connection error by disabling client IP preservation or by disabling cross-zone load balancing.

Additionally, when client IP preservation is enabled, connectivity might fail if the clients that are connecting to the Network Load Balancer are also connected to targets behind the load balancer. To resolve this, you can disable client IP preservation on the affected target groups. Alternatively, have your clients connect only to the Network Load Balancer, or only to the targets, but not both.

TCP connection delays

When both cross-zone load balancing and client IP preservation are enabled, a client connecting to different IPs on the same load balancer may be routed to the same target. If the client uses the same source port for both of these connections, the target will receive what appears to be a duplicate connection, which can lead to connection errors and TCP delays in establishing new connections. You can prevent this type of connection error by disabling cross-zone load balancing. For more information, see Cross-zone load balancing.

Potential failure when the load balancer is being provisioned

One of the reasons a Network Load Balancer could fail when it is being provisioned is if you use an IP address that is already assigned or allocated elsewhere (for example, assigned as a secondary IP address for an EC2 instance). This IP address prevents the load balancer from being set up, and its state is failed. You can resolve this by de-allocating the associated IP address and retrying the creation process.

DNS name resolution contains fewer IP addresses than enabled Availability Zones

Ideally your Network Load Balancer provides one IP address per enabled Availability Zone, when they have at least one healthy host in the Availability Zone. When there are no healthy host in a particular Availability Zone, and cross-zone load balancing is disabled, the IP address of the Network Load Balancer respective of that AZ will be removed from DNS.

For example, suppose your Network Load Balancer has three Availability Zones enabled, all of which have at least one healthy registered target instance.