Deploy your Azure API Management instance to a virtual network - internal mode

APPLIES TO: Developer | Premium

Azure API Management can be deployed (injected) inside an Azure virtual network (VNet) to access backend services within the network. For VNet connectivity options, requirements, and considerations, see:

• Using a virtual network with Azure API Management
• Network resource requirements for API Management injection into a virtual network

This article explains how to set up VNet connectivity for your API Management instance in the internal mode. In this mode, you can only access the following API Management endpoints within a VNet whose access you control.

• The API gateway
• The developer portal
• Direct management
• Git

Use API Management in internal mode to:

• Make APIs hosted in your private datacenter securely accessible by third parties outside of it by using Azure VPN connections or Azure ExpressRoute.
• Enable hybrid cloud scenarios by exposing your cloud-based APIs and on-premises APIs through a common gateway.
• Manage your APIs hosted in multiple geographic locations, using a single gateway endpoint.

For configurations specific to the external mode, where the API Management endpoints are accessible from the public internet, and backend services are located in the network, see Deploy your Azure API Management instance to a virtual network - external mode.

Prerequisites

Review the network resource requirements for API Management injection into a virtual network before you begin.

Some prerequisites differ depending on the version (stv2 or stv1) of the compute platform hosting your API Management instance.

Enable VNet connection

Enable VNet connectivity using the Azure portal (stv2 platform)

1. Go to the Azure portal to find your API management instance. Search for and select API Management services.

2. Choose your API Management instance.

3. Select Network > Virtual network.

4. Select the Internal access type.

5. In the list of locations (regions) where your API Management service is provisioned:

   1. Choose a Location.
   2. Select Virtual network and Subnet.
      • The VNet list is populated with Resource Manager VNets available in your Azure subscriptions, set up in the region you are configuring.

6. Select Apply. The Virtual network page of your API Management instance is updated with your new VNet and subnet choices.

7. Continue configuring VNet settings for the remaining locations of your API Management instance.

8. In the top navigation bar, select Save.

   It can take 15 to 45 minutes to update the API Management instance. The Developer tier has downtime during the process. The Basic and higher SKUs don't have downtime during the process.

After successful deployment, you should see your API Management service's private virtual IP address and public virtual IP address on the Overview blade. For more information about the IP addresses, see Routing in this article.

Enable connectivity using a Resource Manager template (stv2 platform)

• Azure Resource Manager template (API version 2021-08-01 )

  

Enable connectivity using Azure PowerShell cmdlets (stv1 platform)

Create or update an API Management instance in a VNet.

Configure NSG rules

Configure custom network rules in the API Management subnet to filter traffic to and from your API Management instance. We recommend the following minimum NSG rules to ensure proper operation and access to your instance. Review your environment carefully to determine more rules that might be needed.

• For most scenarios, use the indicated service tags instead of service IP addresses to specify network sources and destinations.
• Set the priority of these rules higher than that of the default rules.

DNS configuration

In internal VNet mode, you have to manage your own DNS to enable inbound access to your API Management endpoints.

We recommend:

1. Configure an Azure DNS private zone.
2. Link the Azure DNS private zone to the VNet into which you've deployed your API Management service.

Learn how to set up a private zone in Azure DNS.

Access on default host names

When you create an API Management service (contosointernalvnet, for example), the following endpoints are configured by default:

Access on custom domain names

If you don't want to access the API Management service with the default host names, set up custom domain names for all your endpoints, as shown in the following image:

Configure DNS records

Create records in your DNS server to access the endpoints accessible from within your VNet. Map the endpoint records to the private virtual IP address for your service.

For testing purposes, you might update the hosts file on a virtual machine in a subnet connected to the VNet in which API Management is deployed. Assuming the private virtual IP address for your service is 10.1.0.5, you can map the hosts file as follows. The hosts mapping file is at %SystemDrive%\drivers\etc\hosts (Windows) or /etc/hosts (Linux, macOS).

You can then access all the API Management endpoints from the virtual machine you created.

Routing

The following virtual IP addresses are configured for an API Management instance in an internal virtual network.

The load-balanced public and private IP addresses can be found on the Overview blade in the Azure portal.

For more information and considerations, see IP addresses of Azure API Management.

VIP and DIP addresses

Dynamic IP (DIP) addresses will be assigned to each underlying virtual machine in the service and used to access endpoints and resources in the VNet and in peered VNets. The API Management service's public virtual IP (VIP) address will be used to access public-facing resources.

If IP restriction lists secure resources within the VNet or peered VNets, we recommend specifying the entire subnet range where the API Management service is deployed to grant or restrict access from the service.

Learn more about the recommended subnet size.

Example

If you deploy 1 capacity unit of API Management in the Premium tier in an internal VNet, 3 IP addresses will be used: 1 for the private VIP and one each for the DIPs for two VMs. If you scale out to 4 units, more IPs will be consumed for additional DIPs from the subnet.

If the destination endpoint has allow-listed only a fixed set of DIPs, connection failures will result if you add new units in the future. For this reason and since the subnet is entirely in your control, we recommend allow-listing the entire subnet in the backend.

Force tunnel traffic to on-premises firewall using ExpressRoute or network virtual appliance

Forced tunneling lets you redirect or "force" all internet-bound traffic from your subnet back to on-premises for inspection and auditing. Commonly, you configure and define your own default route (0.0.0.0/0), forcing all traffic from the API Management subnet to flow through an on-premises firewall or to a network virtual appliance. This traffic flow breaks connectivity with API Management, since outbound traffic is either blocked on-premises, or NAT'd to an unrecognizable set of addresses that no longer work with various Azure endpoints. You can solve this issue via the following methods:

• Enable service endpoints on the subnet in which the API Management service is deployed for:

  • Azure SQL (required only in the primary region if the API Management service is deployed to multiple regions)
  • Azure Storage
  • Azure Event Hubs
  • Azure Key Vault (required when API Management is deployed on the stv2 platform)

  By enabling endpoints directly from the API Management subnet to these services, you can use the Microsoft Azure backbone network, providing optimal routing for service traffic. If you use service endpoints with a force tunneled API Management, traffic for the preceding Azure services isn't force tunneled. However, the other API Management service dependency traffic remains force tunneled. Ensure that your firewall or virtual appliance doesn't block this traffic, or the API Management service may not function properly.

  Note

  We strongly recommend enabling service endpoints directly from the API Management subnet to dependent services such as Azure SQL and Azure Storage that support them. However, some organizations may have requirements to force tunnel all traffic from the API Management subnet. In this case, ensure that you configure your firewall or virtual appliance to allow this traffic. You will need to allow the complete IP address range of each dependent service, and keep this configuration up to date when the Azure infrastructure changes. Your API Management service may also experience latency or unexpected timeouts because of the force tunneling of this network traffic.

• All the control plane traffic from the internet to the management endpoint of your API Management service is routed through a specific set of inbound IPs, hosted by API Management, encompassed by the ApiManagement service tag. When the traffic is force tunneled, the responses won't symmetrically map back to these inbound source IPs and connectivity to the management endpoint is lost. To overcome this limitation, configure a user-defined route (UDR) for the ApiManagement service tag with next hop type set to "Internet", to steer traffic back to Azure.

  Note

  Allowing API Management management traffic to bypass an on-premises firewall or network virtual appliance isn't considered a significant security risk. The recommended configuration for your API Management subnet allows inbound management traffic on port 3443 only from the set of Azure IP addresses encompassed by the ApiManagement service tag. The recommended UDR configuration is only for the return path of this Azure traffic.

• (External VNet mode) Data plane traffic for clients attempting to reach the API Management gateway and developer portal from the internet will also be dropped by default because of asymmetric routing introduced by forced tunneling. For each client that requires access, configure an explicit UDR with next hop type "Internet" to bypass the firewall or virtual network appliance.

• For other force tunneled API Management service dependencies, resolve the hostname and reach out to the endpoint. These include:

  • Metrics and Health Monitoring
  • Azure portal diagnostics
  • SMTP relay
  • Developer portal CAPTCHA
  • Azure KMS server

For more information, see Virtual network configuration reference.

Common network configuration issues

This section has moved. See Virtual network configuration reference.

Troubleshooting

Unsuccessful initial deployment of API Management service into a subnet

• Deploy a virtual machine into the same subnet.
• Connect to the virtual machine and validate connectivity to one of each of the following resources in your Azure subscription:
  • Azure Storage blob
  • Azure SQL Database
  • Azure Storage Table
  • Azure Key Vault (for an API Management instance hosted on the stv2 platform)

Verify network status

• After deploying API Management into the subnet, use the portal to check the connectivity of your instance to dependencies, such as Azure Storage.

• In the portal, in the left-hand menu, under Deployment and infrastructure, select Network > Network status.

  

To help troubleshoot connectivity issues, select:

• Metrics - to review network connectivity status metrics

• Diagnose - to run a virtual network verifier over a specified time period

To address connectivity issues, review network configuration settings and fix required network settings.

Incremental updates

When making changes to your network, refer to NetworkStatus API to verify that the API Management service hasn't lost access to critical resources. The connectivity status should be updated every 15 minutes.

To apply a network configuration change to the API Management instance using the portal:

1. In the left-hand menu for your instance, under Deployment and infrastructure, select Network > Virtual network.
2. Select Apply network configuration.

Resource navigation links

An API Management instance hosted on the stv1 compute platform, when deployed into a Resource Manager VNet subnet, reserves the subnet by creating a resource navigation link. If the subnet already contains a resource from a different provider, deployment will fail. Similarly, when you delete an API Management service, or move it to a different subnet, the resource navigation link will be removed.

Challenges encountered in reassigning API Management instance to previous subnet

• VNet lock - When moving an API Management instance back to its original subnet, immediate reassignment may not be possible due to the VNet lock, which takes up to one hour to be removed. If the original subnet has other stv1 platform-based API Management services (cloud service-based), deleting them and waiting is necessary for deploying an stv2 platform-based service in the same subnet.
• Resource group lock - Another scenario to consider is the presence of a scope lock at the resource group level or higher, hindering the Resource Navigation Link Deletion process. To resolve this, remove the scope lock and allow a delay of approximately 4-6 hours for the API Management service to unlink from the original subnet before the lock removal, enabling deployment to the desired subnet.

Troubleshoot connection to Microsoft Graph from inside a VNet

Network connectivity to Microsoft Graph is needed for features including user sign-in to the developer portal using the Microsoft Entra identity provider.

To troubleshoot connectivity to Microsoft Graph from inside a VNet:

• Ensure that NSG and other network rules are configured for outbound connectivity from your API Management instance to Microsoft Graph (using the AzureActiveDirectory service tag).

• Ensure DNS resolution and network access to graph.microsoft.com from within the VNet. For example, provision a new VM inside the VNet, connect to it, and try to GET https://graph.microsoft.com/v1.0/$metadata from a browser or using cURL, PowerShell, or other tools.

Next steps

Learn more about:

• Virtual network configuration reference
• VNet FAQs
• Creating a record in DNS