Troubleshoot Azure Virtual Network NAT connectivity

This article provides guidance on how to troubleshoot and resolve common outbound connectivity issues with your NAT gateway resource. This article also provides best practices on how to design applications to use outbound connections efficiently.

SNAT exhaustion due to NAT gateway configuration

SNAT exhaustion issues with NAT gateway typically have to do with the configurations on the NAT gateway, such as:

  • NAT gateway not scaled out with enough public IP addresses.

  • NAT gateway's configurable TCP idle timeout timer is set higher than the default value of 4 minutes.

NAT gateway not scaled out enough

Each public IP address provides 64,512 SNAT ports for connecting outbound with NAT gateway. From those available SNAT ports, NAT gateway can support up to 50,000 concurrent connections to the same destination endpoint. If outbound connections are dropping because SNAT ports are being exhausted, then NAT gateway may not be scaled out enough to handle the workload. More public IP addresses on NAT gateway may be required in order to provide more SNAT ports for outbound connectivity.

The table below describes two common outbound connectivity failure scenarios due to scalability issues and how to validate and mitigate these issues:

Scenario Evidence Mitigation
You're experiencing contention for SNAT ports and SNAT port exhaustion during periods of high usage. You run the following metrics in Azure Monitor: Total SNAT Connection Count: "Sum" aggregation shows high connection volume. For SNAT Connection Count, "Failed" connection state shows transient or persistent failures over time. Dropped Packets: "Sum" aggregation shows packets dropping consistent with high connection volume and connection failures. Add more public IP addresses or public IP prefixes as need (assign up to 16 IP addresses in total to your NAT gateway). This addition will provide more SNAT port inventory and allow you to scale your scenario further.
You've already assigned 16 IP addresses to your NAT gateway and still are experiencing SNAT port exhaustion. Attempt to add more IP addresses fails. Total number of IP addresses from public IP address or public IP prefix resources exceeds a total of 16. Distribute your application environment across multiple subnets and provide a NAT gateway resource for each subnet.

Note

It is important to understand why SNAT exhaustion occurs. Make sure you are using the right patterns for scalable and reliable scenarios. Adding more SNAT ports to a scenario without understanding the cause of the demand should be a last resort. If you do not understand why your scenario is applying pressure on SNAT port inventory, adding more SNAT ports by adding more IP addresses will only delay the same exhaustion failure as your application scales. You may be masking other inefficiencies and anti-patterns. See best practices for efficient use of outbound connections for additional guidance.

TCP idle timeout timers set higher than the default value

The NAT gateway TCP idle timeout timer is set to 4 minutes by default but is configurable up to 120 minutes. If the timer is set to a higher value than the default, NAT gateway will hold on to flows longer, and can create extra pressure on SNAT port inventory. The table below describes a scenario where a long TCP idle timeout timer is causing SNAT exhaustion and provides mitigation steps to take:

Scenario Evidence Mitigation
You want to ensure that TCP connections stay active for long periods of time without going idle and timing out. You increase the TCP idle timeout timer setting. After a period of time, you start to notice that connection failures occur more often. You suspect that you may be exhausting your inventory of SNAT ports since connections are holding on to them longer. You check the following NAT gateway metrics in Azure Monitor to determine if SNAT port exhaustion is happening: Total SNAT Connection Count: "Sum" aggregation shows high connection volume. For SNAT Connection Count, "Failed" connection state shows transient or persistent failures over time. Dropped Packets: "Sum" aggregation shows packets dropping consistent with high connection volume and connection failures. Some possible steps you can take to resolve SNAT port exhaustion include:

Reduce the TCP idle timeout to a lower value to free up SNAT port inventory earlier. The TCP idle timeout timer can't be set lower than 4 minutes.

Consider asynchronous polling patterns to free up connection resources for other operations.

Use TCP keepalives or application layer keepalives to avoid intermediate systems timing out. For examples, see .NET examples.

Make connections to Azure PaaS services over the Azure backbone using Private Link. This frees up SNAT ports for outbound connections to the internet.

Connection failures due to idle timeouts

TCP idle timeout

As described in the TCP timers section above, TCP keepalives should be used to refresh idle flows and reset the idle timeout. TCP keepalives only need to be enabled from one side of a connection in order to keep a connection alive from both sides. When a TCP keepalive is sent from one side of a connection, the other side automatically sends an ACK packet. The idle timeout timer is then reset on both sides of the connection. To learn more, see Timer considerations.

Note

Increasing the TCP idle timeout is a last resort and may not resolve the root cause. A long timeout can cause low-rate failures when timeout expires and introduce delay and unnecessary failures.

UDP idle timeout

UDP idle timeout timers are set to 4 minutes. Unlike TCP idle timeout timers for NAT gateway, UDP idle timeout timers aren't configurable. The table below describes a common scenario encountered with connections dropping due to UDP traffic idle timing out and steps to take to mitigate the issue.

Scenario Evidence Mitigation
You notice that UDP traffic is dropping connections that need to be maintained for long periods of time. You check the following NAT gateway metrics in Azure Monitor, Dropped Packets: "Sum" aggregation shows packets dropping consistent with high connection volume and connection failures. A few possible mitigation steps that can be taken: - Enable UDP keepalives. Keep in mind that when a UDP keepalive is enabled, it's only active for one direction in a connection. The connection can still go idle and time out on the other side of a connection. To prevent a UDP connection from idle time-out, UDP keepalives should be enabled for both directions in a connection flow. - Application layer keepalives can also be used to refresh idle flows and reset the idle timeout. Check the server side for what options exist for application specific keepalives.

NAT gateway public IP not being used for outbound traffic

VMs hold on to prior SNAT IP with active connection after NAT gateway added to a virtual network

NAT gateway becomes the default route to the internet when configured to a subnet. Migration from default outbound access or load balancer to NAT gateway results in new connections immediately using the IP address(es) associated with the NAT gateway resource. If a virtual machine has an established connection during the migration, the connection will continue to use the old SNAT IP address that was assigned when the connection was established.

Test and resolve issues with VMs holding on to old SNAT IP addresses by:

  • Ensure you've established a new connection and that existing connections aren't being reused in the OS or that the browser is caching the connections. For example, when using curl in PowerShell, make sure to specify the -DisableKeepalive parameter to force a new connection. If you're using a browser, connections may also be pooled.

  • It isn't necessary to reboot a virtual machine in a subnet configured to NAT gateway. However, if a virtual machine is rebooted, the connection state is flushed. When the connection state has been flushed, all connections will begin using the NAT gateway resource's IP address(es). This behavior is a side effect of the virtual machine reboot and not an indicator that a reboot is required.

If you're still having trouble, open a support case for further troubleshooting.

Virtual appliance UDRs and ExpressRoute override NAT gateway for routing outbound traffic

When forced tunneling with a custom UDR is enabled to direct traffic to a virtual appliance or VPN through ExpressRoute, the UDR or ExpressRoute takes precedence over NAT gateway for directing internet bound traffic. To learn more, see custom UDRs.

The order of precedence for internet routing configurations is as follows:
Virtual appliance UDR / ExpressRoute >> NAT gateway >> instance level public IP addresses >> outbound rules on Load balancer >> default outbound access

Test and resolve issues with a virtual appliance UDR or VPN ExpressRoute overriding your NAT gateway by:

  1. Testing that the NAT gateway public IP is used for outbound traffic. If a different IP is being used, it could be because of a custom UDR, follow the remaining steps on how to check for and remove custom UDRs.

  2. Check for UDRs in the virtual network’s route table, refer to view route tables.

  3. Remove the UDR from the route table by following create, change, or delete an Azure route table.

Once the custom UDR is removed from the routing table, the NAT gateway public IP should now take precedence in routing outbound traffic to the internet.

Private Link connects your Azure virtual networks privately to Azure PaaS services such as Azure Storage, Azure SQL, or Azure Cosmos DB over the Azure backbone network instead of over the internet. Private Link uses the private IP addresses of virtual machine instances in your virtual network to connect to these Azure platform services instead of the public IP of NAT gateway. As a result, when looking at the source IP address used to connect to these Azure services, you'll notice that the private IPs of your instances are used. See Azure services listed here for all services supported by Private Link.

To check which Private Endpoints you have set up with Private Link:

  1. From the Azure portal, search for Private Link in the search box.

  2. In the Private Link center, select Private Endpoints or Private Link services to see what configurations have been set up. For more information, see Manage private endpoint connections.

Service endpoints can also be used to connect your virtual network to Azure PaaS services. To check if you have service endpoints configured for your virtual network:

  1. From the Azure portal, navigate to your virtual network and select "Service endpoints" from Settings.

  2. All Service endpoints created will be listed along with which subnets they're configured. For more information, see logging and troubleshooting Service endpoints.

Note

Private Link is the recommended option over Service endpoints for private access to Azure hosted services.

Connection failures at the public internet destination

Connection failures at the internet destination endpoint could be due to multiple possible factors. Factors that can affect connectivity success are:

  • Firewall or other traffic management components at the destination.

  • API rate limiting imposed by the destination side.

  • Volumetric DDoS mitigations or transport layer traffic shaping.

Use NAT gateway metrics in Azure monitor to diagnose connection issues:

  • Look at packet count at the source and the destination (if available) to determine how many connection attempts were made.

  • Look at dropped packets to see how many packets were dropped by NAT gateway.

What else to check for:

  • Check for SNAT exhaustion.

  • Validate connectivity to an endpoint in the same region or elsewhere for comparison.

  • If you're creating high volume or transaction rate testing, explore if reducing the rate reduces the occurrence of failures.

  • If changing rate impacts the rate of failures, check if API rate limits, or other constraints on the destination side might have been reached.

Other transient outbound connectivity issues

Outbound Passive FTP may not work for NAT gateway with multiple public IP addresses, depending on your FTP server configuration.

Passive FTP establishes different connections for control and data channels. When a NAT gateway with multiple public IP addresses sends traffic outbound, it randomly selects one of its public IP addresses for the source IP address. FTP may fail when data and control channels use different source IP addresses, depending on your FTP server configuration.

To prevent possible passive FTP connection failures, make sure to do the following:

  1. Check that your NAT gateway is attached to a single public IP address rather than multiple IP addresses or a prefix.
  2. Make sure that the passive port range from your NAT gateway is allowed to pass any firewalls that may be at the destination endpoint.

Extra network captures

If your investigation is inconclusive, open a support case for further troubleshooting and collect the following information for a quicker resolution. Choose a single virtual machine in your NAT gateway configured subnet to perform the following tests:

  • Use ps ping from one of the backend VMs within the virtual network to test the probe port response (example: ps ping 10.0.0.4:3389) and record results.

  • If no response is received in these ping tests, run a simultaneous Netsh trace on the backend VM, and the virtual network test VM while you run PsPing then stop the Netsh trace.

Outbound connectivity best practices

Azure monitors and operates its infrastructure with great care. However, transient failures can still occur from deployed applications, there's no guarantee that transmissions are lossless. NAT gateway is the preferred option to connect outbound from Azure deployments in order to ensure highly reliable and resilient outbound connectivity. In addition to using NAT gateway to connect outbound, use the guidance below for how to ensure that applications are using connections efficiently.

Modify the application to use connection pooling

When you pool your connections, you avoid opening new network connections for calls to the same address and port. You can implement a connection pooling scheme in your application where requests are internally distributed across a fixed set of connections and reused when possible. This setup constrains the number of SNAT ports in use and creates a predictable environment. Connection pooling helps reduce latency and resource utilization and ultimately improve the performance of your applications.

To learn more on pooling HTTP connections, see Pool HTTP connections with HttpClientFactory.

Modify the application to reuse connections

Rather than generating individual, atomic TCP connections for each request, configure your application to reuse connections. Connection reuse results in more performant TCP transactions and is especially relevant for protocols like HTTP/1.1, where connection reuse is the default. This reuse applies to other protocols that use HTTP as their transport such as REST.

Modify the application to use less aggressive retry logic

When SNAT ports are exhausted or application failures occur, aggressive or brute force retries without delay and back-off logic cause exhaustion to occur or persist. You can reduce demand for SNAT ports by using a less aggressive retry logic.

Depending on the configured idle timeout, if retries are too aggressive, connections may not have enough time to close and release SNAT ports for reuse.

For extra guidance and examples, see Retry pattern.

Use keepalives to reset the outbound idle timeout

For more information about keepalives, see TCP idle timeout timers set higher than the default value.

When possible, Private Link should be used to connect directly from your virtual networks to Azure platform services in order to reduce the demand on SNAT ports. Reducing the demand on SNAT ports can help reduce the risk of SNAT port exhaustion.

To create a Private Link, see the following Quickstart guides to get started:

Next steps

We're always looking to improve the experience of our customers. If you're experiencing issues with NAT gateway that aren't listed or resolved by this article, submit feedback through GitHub via the bottom of this page. We'll address your feedback as soon as possible.

To learn more about NAT gateway, see: