Domain 4 of 5 · Chapter 1 of 3

Configure and manage virtual networks in Azure

Virtual networks and the subnet address plan

Picture a virtual network created in East US with the address space 10.1.0.0/16 and a subnet 10.1.1.0/24 for application VMs: before the first VM arrives, five of that subnet's 256 addresses are already taken by Azure. Sizing that subnet, peering its virtual network, giving VMs public IP addresses, steering traffic with custom (user-defined) routes and troubleshooting a failed connection are the tasks the Microsoft AZ-104 study guide skills measured lists under 'Configure and manage virtual networks in Azure' (Study guide for Exam AZ-104[1]). The Microsoft AZ-104 study guide skills measured lists 'Evaluate effective security rules in NSGs' (NSGs are network security groups), 'Implement Azure Bastion', 'Configure service endpoints for Azure platform as a service (PaaS)' and 'Configure private endpoints for Azure PaaS' under 'Configure secure access to virtual networks' (Implement and manage virtual networking) (Study guide for Exam AZ-104[1]); see Configure secure access to virtual networks. It lists 'Configure Azure DNS', 'Configure an internal or public load balancer' and 'Troubleshoot load balancing' under 'Configure name resolution and load balancing' (Implement and manage virtual networking) (Study guide for Exam AZ-104[1]); see Configure name resolution and load balancing.

A virtual network is one region's private address space

A virtual network is a logical isolation of the Azure cloud dedicated to your subscription; it has its own CIDR (Classless Inter-Domain Routing) address block, which you divide into subnets (Azure Virtual Network FAQ[2]). It is limited to a single region but spans that region's availability zones, so resources in different zones of one region can share it, while a second region needs a second virtual network, connected to the first through peering when the two must communicate (Azure Virtual Network FAQ[2]).

Microsoft recommends the RFC 1918 private ranges (10.0.0.0/8, 172.16.0.0/12 and 192.168.0.0/16) and also treats the RFC 6598 shared range 100.64.0.0/10 as private space (Azure Virtual Network FAQ[2]). A publicly routable range is allowed inside a virtual network, but it does not make those resources reachable from the internet: Microsoft states that such addresses aren't directly accessible from the internet, and that ranges outside the recommended ones might have undesirable side effects, so private addressing is the normal choice (Azure Virtual Network FAQ[2]). Choose a space that does not overlap any network you might connect later, because overlapping CIDR blocks prevent peering and require address translation for on-premises VPN connectivity (Azure Virtual Network FAQ[2], NAT on Azure VPN Gateway[3]).

Subnets: inside the space, never overlapping, five addresses reserved

A subnet's range must be unique within the virtual network's address space and can't overlap another subnet's range (Add, change, or delete a virtual network subnet[4]). Azure reserves the first four addresses and the last address of every subnet (Azure Virtual Network FAQ[2]):

Address in 192.168.1.0/24 Reserved for
192.168.1.0 Network address
192.168.1.1 The default gateway
192.168.1.2 and 192.168.1.3 Mapping the Azure DNS IP addresses into the virtual network
192.168.1.255 Network broadcast address

A /24 therefore leaves 251 assignable addresses: the first address a resource can receive in 10.1.1.0/24 is 10.1.1.4, and in a second subnet, 10.1.2.0/24, it is 10.1.2.4 (Azure Virtual Network FAQ[2]). The smallest supported IPv4 subnet is /29, the largest is /2, and an IPv6 subnet must be exactly /64 (Azure Virtual Network FAQ[2]); a /29 holds eight addresses, of which three are usable (Add, change, or delete a virtual network subnet[4]). Size each subnet for the addresses that remain after the reservation, plus any minimum a service imposes on its own subnet (the VPN gateway subnet described later in this topic is one example).

The figure below shows the containment this section describes: a region holds the virtual network, the virtual network's address space holds non-overlapping subnets, and each VM's network interface (NIC), the resource that connects a VM to a subnet, takes its private address from one subnet.

The takeaway: plan the address space once, with room to grow and no overlap with anything you may connect later, because peering, routes and outbound access are all expressed in these prefixes.

Region: East USVirtual network10.1.0.0/16Subnet10.1.1.0/24251 usableSubnet10.1.2.0/24251 usableNIC of VM110.1.1.4NIC of VM210.1.2.4Reserved in every subnet:first four and last address
Address containment in one region: virtual network address space, two non-overlapping subnets, and the NIC addresses taken from them

Creating and changing subnets

Subnets change over a network's life: you add them, grow them, hand them to a service and delete them, and Azure protects any subnet that still holds resources.

Creating a subnet: the Azure CLI writes directly, Azure PowerShell needs a save

The two command-line tools take different approaches. The Azure CLI acts on the named virtual network in one call: az network vnet subnet create and az network vnet subnet delete identify their target with --resource-group and --vnet-name (Add, change, or delete a virtual network subnet[4]). Azure PowerShell works on an object instead: Get-AzVirtualNetwork retrieves the virtual network into a variable, Add-AzVirtualNetworkSubnetConfig adds the subnet configuration to that object, and Set-AzVirtualNetwork then applies the object to the virtual network in Azure; deleting follows the same pattern, with Remove-AzVirtualNetworkSubnetConfig piped to Set-AzVirtualNetwork (Add, change, or delete a virtual network subnet[4]). Until Set-AzVirtualNetwork runs, the change exists only in the object in your PowerShell session.

Listing: adding an example subnet, snet-web (10.1.1.0/24), to an example virtual network, vnet-app, first with Azure PowerShell, then with the Azure CLI

# Retrieve the virtual network into a local object
$vnet = Get-AzVirtualNetwork -Name 'vnet-app' -ResourceGroupName 'rg-network'
# Add the subnet configuration to the object (Azure is not changed yet)
Add-AzVirtualNetworkSubnetConfig -Name 'snet-web' -VirtualNetwork $vnet -AddressPrefix '10.1.1.0/24'
# Apply the modified object to the virtual network in Azure
Set-AzVirtualNetwork -VirtualNetwork $vnet
# One call; --resource-group and --vnet-name identify the target network
az network vnet subnet create --resource-group rg-network --vnet-name vnet-app \
  --name snet-web --address-prefixes 10.1.1.0/24

Changing the range of a subnet that holds resources

You can change a subnet's address range only while no resources are deployed in it; otherwise you must first move those resources to another subnet or delete them, and you can delete a subnet only when it contains no resources (Add, change, or delete a virtual network subnet[4]). The multiple-prefixes capability is the exception for VM workloads: a subnet of VMs or VM scale sets can gain another address prefix without first removing its resources, and when the first prefix is full, new VMs spill over into the added one (Create multiple prefixes for a subnet[5]). Its limits are part of the decision (Create multiple prefixes for a subnet[5]):

  • It supports only virtual machines and virtual machine scale sets. Delegated subnets can't use it, except a GatewaySubnet delegated to ExpressRoute Gateway services. GatewaySubnet is the dedicated subnet a virtual network gateway is deployed into (see the VPN Gateway section below), and ExpressRoute is a private connection, through a connectivity provider, between an on-premises network and the Microsoft cloud that doesn't go over the public internet; its gateway is a different gateway type from a VPN gateway (About VPN Gateway configuration settings[6], What is Azure ExpressRoute?[7]).
  • Each NIC in such a subnet supports a single IPv4 and a single IPv6 address.
  • You configure it with Azure PowerShell, the Azure CLI or Azure Resource Manager (ARM) templates; portal support is limited.
  • An update applies exactly the prefixes you list: any prefix you omit is removed if unused, and causes an error if a NIC still references it. List every prefix you want to keep.

Microsoft also advises using only the addressPrefixes property (a list) for both single and multiple prefixes, never mixing it with addressPrefix (Create multiple prefixes for a subnet[5]). That caution names two subnet properties; the -AddressPrefix in the PowerShell listing above is a cmdlet parameter that takes a list of prefixes (String[]) (Add-AzVirtualNetworkSubnetConfig[8]), and the same Microsoft article passes one or several prefixes through -AddressPrefix in PowerShell and through --address-prefixes in the Azure CLI (Create multiple prefixes for a subnet[5]). Outside this capability, replacing a range that is in use still means moving or removing the resources first.

Subnet delegation

Subnet delegation gives an Azure service explicit permission to create its service-specific resources in a subnet, and a subnet can carry zero or more delegations (Add, change, or delete a virtual network subnet[4]). Configure the delegation that the service you are deploying requires. Once a service's resources are deployed in the subnet, you can't add or remove delegations until you remove all of that service's resources (Add, change, or delete a virtual network subnet[4]). Delegation is a different feature from a service endpoint: a service endpoint concerns how resources in a subnet reach a PaaS service, and it is covered in Configure secure access to virtual networks.

In short: empty subnets are easy to change, populated ones change only by moving resources out or, for VM subnets, by adding prefixes, and a PowerShell edit takes effect only after Set-AzVirtualNetwork.

Network interfaces: how a VM attaches to subnets

A VM never sits in a subnet itself: its network interfaces do, so the rules for NICs decide which subnets a VM can join directly and where a NIC may live.

The placement boundary: same subscription and region

A NIC can be in the same resource group as its VM and its virtual network or in a different one. It can be assigned only to a virtual network in the same subscription and region as the NIC, and the VM it is added to must be in that same subscription and region (Create, change, or delete a network interface[9]). After creation you can change a NIC's subnet but not its virtual network; to move it to another subnet, first set all its private IP addresses to dynamic assignment (Create, change, or delete a network interface[9]). A VM therefore can't attach a NIC from another region: virtual network peering, described later, gives routed connectivity between networks, not a way to attach an interface from elsewhere.

One NIC per subnet the VM must join directly

When a VM must connect directly to two subnets, attach one NIC for each. All NICs of a VM must connect to the same virtual network, although they can be in different subnets of it, and the VM size limits how many NICs it can carry (Add network interfaces to or remove network interfaces from virtual machines[10]). Adding or removing a NIC on an existing VM requires the VM to be stopped (deallocated), and afterwards you configure the guest operating system to use the extra interfaces (Add network interfaces to or remove network interfaces from virtual machines[10]). The first NIC attached is the primary one, and by default the VM sends all outbound traffic to the address of the primary IP configuration of the primary NIC (Add network interfaces to or remove network interfaces from virtual machines[10]). The portal creates a VM with a single NIC and can't attach existing NICs to a new VM; use the Azure CLI or PowerShell for that (Add network interfaces to or remove network interfaces from virtual machines[10]). A network bridge configured inside the guest is no substitute: virtual networks are Layer 3 overlays without Layer 2 semantics (Azure Virtual Network FAQ[2]), so only an Azure NIC places the VM in a subnet.

IP configurations: more addresses on the same subnet

Every NIC has one or more IP configurations; an ordinary single-address configuration holds one static or dynamic private IP address and can have one public IP address resource associated with it (Assign multiple IP addresses to virtual machines[11]). The first is the primary IP configuration and any others are secondary. All IP configurations on a single NIC must be associated with the same subnet, so extra configurations give a VM more addresses on that subnet, never a presence in another one; addresses in different subnets need multiple NICs (Assign multiple IP addresses to virtual machines[11]). After adding secondary private addresses, add them inside the guest operating system as well, but never configure the VM's public IP address inside the guest (Assign multiple IP addresses to virtual machines[11]).

IPv6 is dual-stack only

Azure doesn't support IPv6-only VMs or VM scale sets: each NIC must include at least one IPv4 IP configuration (What is IPv6 for Azure Virtual Network?[12]). To give a NIC an IPv6 address, the virtual network and subnet must also have IPv6 address space, and the IPv6 address is assigned to a secondary IP configuration (Create, change, or delete a network interface[9]).

Accelerated Networking

Accelerated Networking enables single root I/O virtualization (SR-IOV) on supported VM types: traffic arriving at the NIC is forwarded directly to the VM, bypassing the host and its virtual switch, which lowers latency, jitter and CPU utilization (Azure Accelerated Networking overview[13]). It doesn't bypass your network policy: the policies the virtual switch applied, including network security groups, access control lists and isolation, are offloaded to hardware and still enforced (Azure Accelerated Networking overview[13]). Confirm that the VM type and operating system are supported, and stop and deallocate an existing VM before enabling it, because it can't be enabled on a running VM (Azure Accelerated Networking overview[13]).

The NIC is the unit of attachment: its subnet decides its addresses, the number of NICs decides how many subnets a VM joins directly, and its subscription and region must match the VM's.

Private IP address allocation

Every IP configuration's private address comes from Azure's allocation on the NIC, and the guest operating system only receives it, so allocation is a NIC setting, not a guest setting.

Azure offers two allocation methods (Private IP addresses[14]):

Method Who chooses the address When the address is released or changes
Dynamic (the default) Azure assigns an available address from the subnet; normally the next in sequence, but that isn't guaranteed The NIC is deleted, moved to a different subnet in the same virtual network, or switched to static with a different address specified
Static You select any unassigned, unreserved address in the subnet The NIC is deleted

Two consequences matter for administrators. First, stopping and deallocating a VM doesn't release a dynamic address: for VMs deployed through ARM, private addresses don't change whether they were assigned by the static or the dynamic method (Azure Virtual Network FAQ[2]). Second, switching from dynamic to static without specifying another address keeps the current address by default, so fixing an address in place never renumbers the VM (Private IP addresses[14]). When something depends on one exact address, choose static; you can create a NIC with a static private address before creating its VM (Create, change, or delete a network interface[9]).

Set the address on the Azure NIC, not only inside the guest. The Azure DHCP server assigns the private address to the NIC in the VM's operating system (Create, change, or delete a network interface[9]), and Microsoft doesn't recommend assigning addresses manually inside the operating system unless necessary: if the address configured in the guest differs from the address assigned to the Azure NIC, you lose connectivity to the VM (Azure Virtual Network FAQ[2]). Any guest configuration you do need must match Azure's.

The practical rule: dynamic addresses already survive stop and start, static is for addresses others depend on, and either way the change is made on the NIC.

Public IP address resources and their properties

A public IP address in ARM is a resource with its own properties: you create it, then associate it with the resource that needs internet reachability (Public IP addresses[15]).

SKU and allocation

Basic SKU public IPs were retired on September 30, 2025, and Standard (v1 or v2) is recommended for new ARM-native deployments that do not depend on Cloud Services Extended Support (Public IP addresses[15]). Standard has two versions, v1 and v2, which share static allocation and the closed-to-inbound default but not their consumers: a Standard v2 IP can be used only with the Standard v2 NAT gateway, not with any other product or component, so associating a public IP with a VM's NIC, a load balancer frontend or any consumer listed below other than the Standard v2 NAT gateway is a Standard v1 use (Public IP addresses[15]). The two versions also differ in zone options, Internet routing preference and the Global tier, each described below (Public IP addresses[15]). A Standard public IPv4 or IPv6 address uses static allocation: Azure assigns it from the region's pool when the resource is created, you can't choose the actual address, and it is released only when the public IP resource is deleted, after which it can't be recovered (Public IP addresses[15]). Static addresses aren't released until the public IP resource is deleted, so stopping or deallocating the associated VM releases nothing (Create, change, or delete an Azure public IP address[16]). An address created from a public IP prefix follows the prefix's reservation instead, as the next section explains.

Associating it with a VM

For direct internet connectivity to a VM, associate a Standard v1 public IP with an IP configuration of the VM's NIC; the same resource type also serves public load balancer frontends, virtual network gateways, NAT gateways, Application Gateway, Azure Firewall, Bastion hosts and other resources (Public IP addresses[15]). It must be in the same subscription and region as the resource it is associated with, but it can be in a different resource group (Create, change, or delete an Azure public IP address[16]).

Standard public IPs follow a secure-by-default model: closed to inbound traffic when used as a frontend, so a network security group (NSG) that allows the traffic is required, for example on the NIC of the VM (Public IP addresses[15]). The address supplies reachability, not an open application port; writing the NSG rule is covered in Configure secure access to virtual networks.

Availability zones

In regions with availability zones, a Standard v1 public IP is zonal or zone-redundant: a zonal IP is tied to one zone and shares fate with that zone's health, while a zone-redundant IP survives any single zone failure (Public IP addresses[15]). Formerly non-zonal Standard v1 IPs are now zone-redundant in every region that supports zones, Standard v2 IPs are always zone-redundant there, and a public IP can't change its availability zone after creation (Public IP addresses[15]). The setting protects only the address frontend: a backend VM running in one zone still fails with that zone, and placing VMs in an availability set doesn't change the IP's zone configuration.

Tier and SKU must match the consumer

The tier is Regional or Global; a Global tier IP is an anycast address used with cross-region load balancers, and Standard v2 doesn't yet support that tier (Public IP addresses[15]). A load balancer and its public IP need matching SKUs (Public IP addresses[15]), and a public IP that serves a Standard load balancer must align with the load balancer's tier (Upgrade Basic public IP addresses to Standard SKU[17]), so a regional Standard load balancer takes a Regional Standard v1 IP. Choosing a Global tier IP doesn't by itself make a regional load balancer span regions; that needs a cross-region load balancer. A Standard v2 IP fits no load balancer, regional or cross-region, because it serves only the Standard v2 NAT gateway (see SKU and allocation above). Pick the SKU and tier the consuming service supports, not just any resource labeled Standard.

DNS name label

A DNS name label maps label.region.cloudapp.azure.com to the public IP in Azure-managed DNS (contoso in West US becomes contoso.westus.cloudapp.azure.com), as an A record for IPv4 or an AAAA record for IPv6, and each label must be unique within its Azure region (Public IP addresses[15]). For a name in your organization's own domain, create the record in Azure DNS or with an external DNS provider instead (Public IP addresses[15]); public DNS zones are covered in Configure name resolution and load balancing.

Routing preference

Routing preference decides how traffic travels between Azure and the internet (What is routing preference?[18]). The default, Microsoft network, keeps traffic on the Microsoft global network for most of its journey (cold potato routing); Internet hands traffic to the transit ISP network near the Azure region (hot potato routing), and the egress data transfer price differs between the two (What is routing preference?[18]). You choose it when creating the public IP and can't change it afterwards (What is routing preference?[18]). Internet preference works only with zone-redundant Standard v1 public IPs (not Standard v2 or Basic), supports only IPv4, and can't be used with NAT gateways or IP-based public load balancers; with it, traffic to destinations within Azure still takes the direct path within the Microsoft network (What is routing preference?[18]). Routing preference is a property of the public IP; it is unrelated to a user-defined route whose next hop type is Internet (route types are covered later in this topic).

To sum up: a Standard public IP is a static, closed-by-default resource, region-bound when its tier is Regional, whose zone and routing preference are fixed at creation, and whose SKU and tier must fit the service that uses it.

Public IP lifecycle: prefixes, moves and Basic upgrades

Public IP addresses often outlive the VMs they serve, so their lifecycle has its own rules for reserving ranges, moving and deleting resources, and upgrading the retired Basic SKU.

Public IP prefix: a reserved contiguous range

When an external firewall must allowlist your addresses, allocate them from a public IP prefix, a reserved range of public IP addresses in one region and subscription (Public IP address prefix[19]). One firewall rule then covers the whole prefix, and VMs you add as you scale take addresses from the same range (Public IP address prefix[19]).

IPv4 prefix IPv6 prefix Addresses
/28 /124 16
/29 /125 8
/30 /126 4
/31 /127 2

These are the available sizes for Microsoft-owned ranges (a prefix derived from your own custom IP prefix can be larger), and the size can't be modified after the prefix is created (Public IP address prefix[19]). Only static Standard SKU public IPs can be assigned from a prefix, and for Standard v2 you create a v2 prefix and v2 IPs from it for a Standard v2 NAT gateway (Public IP address prefix[19]). The lifecycle differs from an individually allocated public IP: deleting a public IP created from a prefix returns its address to your reserved range for later reuse, and the range stays reserved until you delete the prefix (Public IP address prefix[19]). A prefix is your own reservation; it is unrelated to a service tag, which names the address ranges of an Azure service.

Moving and deleting

Standard public IPs can move under these conditions (Create, change, or delete an Azure public IP address[16]):

Move Supported Condition
Across resource groups Yes While attached to any resource except a virtual network gateway
Across subscriptions Yes Only when detached from all resources
Across regions No IP addresses are tied to regional ranges

Deleting a public IP requires that it isn't associated with any IP configuration or VM network interface, so dissociate it first; once deleted, the address can't be recovered (Create, change, or delete an Azure public IP address[16]). A reorganization that must keep the address therefore uses a supported move, never deletion and recreation.

Upgrading an existing Basic public IP

Basic public IPs keep working after their September 30, 2025 retirement, but they are unsupported and not covered by SLA guarantees (Upgrade Basic public IP addresses to Standard SKU[17]). Microsoft's guidance is to identify the resource consuming each Basic IP, create a migration plan for planned downtime, and then follow the path for that resource (Upgrade Basic public IP addresses to Standard SKU[17]):

Resource using the Basic IP Upgrade path
Virtual machine Detach and upgrade the public IPs, manually or with Microsoft's upgrade scripts
Basic Load Balancer A new load balancer SKU: upgrade to Standard Load Balancer with the upgrade script
VPN gateway VPN Gateway migration
ExpressRoute gateway The ExpressRoute gateway migration experience

The table is abridged; the guidance also covers scale sets, Application Gateway v1 and Azure Databricks.

A standalone public IP upgrade has three conditions (Upgrade a public IP address[20]): the Basic IP must use static assignment, it must not be associated with any resource, and the upgrade can't be reversed. Upgrading the resource retains its IP address (Upgrade a public IP address[20]). Because removing a public IP with dynamic assignment from its resource releases the address (Create, change, or delete an Azure public IP address[16]), switch a dynamic Basic IP to static before you detach it whenever the address must survive. Standard IPs are secure by default, so apply an NSG to the associated NIC or subnet to allow traffic and prevent unexpected downtime (Upgrade a public IP address[20]).

The rule across this section: an address lives as long as its public IP resource (or its prefix), so keep it by moving or upgrading the resource, never by deleting and recreating it.

Virtual network peering: requirements, links and reach

Virtual network peering is how two Azure virtual networks reach each other privately, and it can connect spokes to a hub: each spoke virtual network peers with one hub virtual network that hosts shared infrastructure such as a VPN gateway (Azure virtual network peering[21]).

What peering provides

Peering connects virtual networks in the same region (virtual network peering) or in different regions (global virtual network peering), and it works across subscriptions, Microsoft Entra tenants, deployment models and regions (Azure virtual network peering[21]). Traffic between peered networks is private and stays on the Microsoft backbone; no public internet, gateways or encryption are required, and creating the peering causes no downtime (Azure virtual network peering[21]). Within one region the latency matches a single virtual network, and the peering adds no bandwidth restriction beyond the VM's own (Azure virtual network peering[21]). The networks you peer must have nonoverlapping address spaces (Create, change, or delete a virtual network peering[22]). A private endpoint is not an alternative: it exposes one service, not a whole network (see Configure secure access to virtual networks).

A peering is two links

A peering exists as a link in each of the two virtual networks. Creating the first link sets its status to Initiated; creating the second sets both to Connected, and the peering isn't established until both links show Connected (Create, change, or delete a virtual network peering[22]). If one link is later deleted, the peering goes Disconnected; delete the remaining link and recreate both (Azure Virtual Network FAQ[2]). The portal's add-peering page collects settings for both the local and the remote link, while Add-AzVirtualNetworkPeering in PowerShell and az network vnet peering create in the Azure CLI each create one direction, so a scripted peering runs once per network (Create, change, or delete a virtual network peering[22]).

Listing: peering vnet-1 and vnet-2 with the Azure CLI, one command per direction

# Link 1: vnet-1 to vnet-2 (status Initiated until link 2 exists)
az network vnet peering create --name vnet-1-to-vnet-2 \
  --vnet-name vnet-1 --remote-vnet vnet-2 \
  --resource-group rg-network --allow-vnet-access
# Link 2: vnet-2 back to vnet-1 (both links now show Connected)
az network vnet peering create --name vnet-2-to-vnet-1 \
  --vnet-name vnet-2 --remote-vnet vnet-1 \
  --resource-group rg-network --allow-vnet-access

--allow-vnet-access sets whether the local network's VMs can access the remote network's VMs (az network vnet peering[23]), the same choice as the portal setting Allow access to the peered virtual network (Create, change, or delete a virtual network peering[22]), listed under peering settings below. Both example networks sit in one resource group. When the networks are in different subscriptions or different resource groups, pass the full resource ID of the remote virtual network to PowerShell or the CLI (Create, change, or delete a virtual network peering[22]).

Permissions, including across subscriptions and tenants

The accounts that work with peerings need the Network Contributor role, or a custom role with the actions Microsoft.Network/virtualNetworks/virtualNetworkPeerings/write, .../read and .../delete; removing the remote network's link requires those permissions on the remote virtual network (Create, change, or delete a virtual network peering[22]). If you lack read access to the remote network, so the portal can't list it, the I know my resource ID option lets you enter its resource ID (Create, change, or delete a virtual network peering[22]). The ID only identifies the remote network and grants no right to change it, so when one account doesn't control both networks, coordinate with the remote administrator to create the reciprocal link. Resource groups play no part in eligibility, and a virtual network can't even be moved to another resource group or subscription while it is part of a peering: delete the peering, move the network, then recreate the peering (Create, change, or delete a virtual network peering[22]).

Peering is not transitive

Peerings by themselves aren't transitive: with peerings between VNet1 and VNet2 and between VNet2 and VNet3, there is no connectivity between VNet1 and VNet3 through VNet2 (Create, change, or delete a virtual network peering[22]). For direct communication, create an explicit peering between VNet1 and VNet3, or send the traffic through a network virtual appliance (NVA) in the hub network (Create, change, or delete a virtual network peering[22]), a VM such as a firewall that forwards traffic not addressed to itself, with user-defined routes (UDRs, custom routes described under route tables below) whose next hop is that VM (Azure virtual network peering[21]); the section on routing through a network virtual appliance below covers its setup. Name resolution doesn't cross a peering either: default Azure name resolution can't resolve names in peered networks, so use Azure Private DNS or a custom DNS server (Create, change, or delete a virtual network peering[22]), covered in Configure name resolution and load balancing.

The figure below shows the three networks: VNet2 is peered with both others, yet VNet1 has no path to VNet3 until one of the two fixes is in place.

Resizing a peered network

You can resize the address space of a peered virtual network without downtime on the currently peered ranges, by modifying a prefix, adding a range or deleting one; afterwards the peers must sync with the new address space, through the portal or Azure PowerShell, and Microsoft recommends syncing after every resize operation (Azure virtual network peering[21]). The feature doesn't support a virtual network peered with a classic virtual network (Azure virtual network peering[21]). When workloads in a newly added range are unreachable across an otherwise Connected peering, check the sync status and the effective routes: by default every subnet of a peered network has a route with next hop type Virtual network peering for each address space of the peer (Azure virtual network peering[21]).

The takeaway: a peering needs nonoverlapping ranges, two Connected links and permissions on each side, and it connects exactly two networks.

VNet2VNet1VNet3PeeredPeeredNo pathFixes for VNet1 to VNet3:Peer VNet1and VNet3directlyRoute throughan NVA in VNet2with UDRs
Non-transitive peering: VNet2 peers with VNet1 and VNet3, VNet1 still cannot reach VNet3, and the two ways to connect them

VPN Gateway: site-to-site and point-to-site connections

When the far end of a connection is not an Azure virtual network but an on-premises site or a single remote computer, peering doesn't apply; Azure VPN Gateway does.

Azure VPN Gateway sends encrypted traffic between an Azure virtual network and on-premises locations over the public internet, using a specific type of virtual network gateway called a VPN gateway (What is Azure VPN Gateway?[24]). It offers two connection types for that job (What is Azure VPN Gateway?[24]):

Connection Connects Tunnel
Site-to-site An on-premises network, through its VPN device IPsec/IKE
Point-to-site One remote computer, for example from home or a conference OpenVPN, IKEv2 or SSTP

Choose by what sits at the far end: a whole site with a VPN device needs site-to-site, while an individual user's computer needs point-to-site and no site VPN device. VPN Gateway also supports VNet-to-VNet connections between two Azure VPN gateways (What is Azure VPN Gateway?[24]).

The gateway is deployed into a dedicated subnet that must be named GatewaySubnet; most gateway SKUs need it to be /27 or larger, and nothing else may be deployed into it (About VPN Gateway configuration settings[6]). A peered virtual network can share this gateway instead of deploying its own, through gateway transit (next section).

A public IP on a VM is not an alternative to a VPN: it supplies an internet-reachable address, but neither an encrypted tunnel nor routes into the private network. When the only requirement is administrator access to VMs, Azure Bastion is the fit: it provides RDP and SSH connectivity to VMs over TLS, reaching them by their private IP addresses so they need no public IP (What is Azure Bastion?[25]). Bastion gives sessions on VMs, not general network access for a remote user's applications; its deployment is covered in Configure secure access to virtual networks.

In one line: site-to-site for networks, point-to-site for individual computers, Bastion for VM administration only.

Peering settings: forwarded traffic and gateway transit

Beyond direct VM-to-VM traffic, each peering link carries settings that decide what else the link accepts; two of them, forwarded traffic and gateway transit, are separate features that are easily confused.

The portal shows these settings for each link, and all except the first are off by default (Create, change, or delete a virtual network peering[22]):

Setting on a link Default What it allows
Allow access to the peered virtual network On VMs in the two networks to communicate; the VirtualNetwork service tag then includes the peered network
Allow forwarded traffic Off This network to receive traffic that came from beyond the peer and was forwarded through it, for example by an NVA in the peer
Allow gateway transit Off The peer to receive traffic from this network's gateway or Route Server; requires a gateway or Route Server in this network
Use remote gateways Off This network to use the peer's gateway or Route Server; requires the peer to have one and to allow gateway transit

The portal's labels are longer (for example, Allow gateway or route server in vnet-1 to forward traffic to the peered virtual network); this topic uses the short names in the first column.

Forwarded traffic

Allow forwarded traffic admits traffic that an appliance in the peer forwarded from elsewhere. For example, if an NVA in VNet2 receives traffic from outside VNet2 and forwards it to VNet1, enabling the setting on VNet1's link lets that traffic reach VNet1 (Create, change, or delete a virtual network peering[22]). It is separate from Allow access and creates no user-defined routes (UDRs, the custom routes described under route tables below) and no NVA; those are configured separately (Create, change, or delete a virtual network peering[22]), including the routes for the return direction.

Gateway transit

Gateway transit lets a peered virtual network use a VPN or ExpressRoute gateway in another network instead of deploying its own (Azure virtual network peering[21]). It takes a setting on both links: Allow gateway transit on the link of the network that owns the gateway (the hub), and Use remote gateways on the link of the network that consumes it (the spoke) (Create, change, or delete a virtual network peering[22]). Its constraints:

The two features don't substitute for each other: Allow forwarded traffic never selects a hub gateway, so it can't give a spoke shared VPN connectivity, and Use remote gateways never admits traffic that a firewall VM forwards. Gateway transit also doesn't make peering transitive for spoke-to-spoke traffic forwarded by a hub VM; that still needs a direct peering or an NVA with UDRs, as the section on virtual network peering above describes.

Remember which link carries which setting: Allow forwarded traffic on the link of the receiving network, Allow gateway transit on the hub's link, and Use remote gateways on the spoke's link.

System routes and how Azure selects a route

Every packet that leaves a subnet is matched against that subnet's routes, and most of those routes exist before you configure anything: Azure automatically creates system routes and assigns them to each subnet in a virtual network (Virtual network traffic routing[26]).

Default system routes

You can't create or remove system routes, but you can override some of them with custom routes (Virtual network traffic routing[26]). The defaults include the following (Virtual network traffic routing[26]):

Address prefix Next hop type
Each address range of the virtual network Virtual network
0.0.0.0/0 Internet
10.0.0.0/8, 172.16.0.0/12 and 192.168.0.0/16 (RFC 1918), 100.64.0.0/10 (RFC 6598) None

The table is abridged: Azure also adds next hop None routes for several other reserved prefixes. What each next hop type means (Virtual network traffic routing[26]):

  • Next hop Virtual network carries traffic between the address ranges of the virtual network, so Azure routes between subnets without any route table or gateway, VPN or otherwise.
  • Next hop Internet handles every destination no other route covers. Traffic to the public IP addresses of Azure services stays on the Azure backbone rather than crossing the internet.
  • Next hop None drops the traffic. When you add one of those private ranges to the virtual network's address space, Azure changes that route's next hop to Virtual network.

A route chooses a path; it doesn't translate addresses. The Internet route alone doesn't give a VM in a private subnet (one with default outbound access disabled) a public source address (see the section on outbound internet access below).

Azure adds optional system routes when you enable features: next hop Virtual network peering for each address range of a peered network, next hop Virtual network gateway for prefixes learned from on-premises through Border Gateway Protocol (BGP) or configured in a local network gateway, and VirtualNetworkServiceEndpoint on subnets with a service endpoint (Virtual network traffic routing[26]).

Route selection: longest prefix first

Azure selects a route for outbound traffic by the destination address, using the longest prefix match (Virtual network traffic routing[26]). With routes for 10.0.0.0/24 and 10.0.0.0/16, traffic to 10.0.0.5 takes the /24 route and traffic to 10.0.1.5 takes the /16 route (Virtual network traffic routing[26]). Only when several routes have the same prefix does the source decide, in this order: user-defined route (UDR), then BGP route, then system route (Virtual network traffic routing[26]). Unlike an NSG, which processes its rules in priority order and stops at the first match (Azure network security groups overview[27]), a route wins by prefix length; see 'How NSG rules are evaluated' in Configure secure access to virtual networks.

Two exceptions apply: system routes for the virtual network, for virtual network peerings and for service endpoints are preferred even when a BGP route is more specific (overriding longest prefix match), and routes with a service endpoint as next hop can't be overridden even with a route table (Virtual network traffic routing[26]). In practice, a broad 0.0.0.0/0 UDR doesn't capture traffic to a peered network, because the peering route has the longer prefix.

The figure below walks the selection for one destination.

Read routes by prefix length first; the source of a route breaks only a tie, apart from the preferred system routes above.

Packet leaves a subnetfor one destinationKeep routes whose prefixcontains the destinationOne longestmatching prefix?YesNo (tie)Use that routeTie: UDR first,then BGP,then system routeException: system routes forthe virtual network, peeringsand service endpoints beateven more specific BGP routes
Azure route selection for one destination: longest prefix match, then route source (UDR, BGP, system) on a tie, with the preferred system route exception

Route tables and user-defined routes

To send traffic somewhere other than where the system routes would, you write user-defined routes (UDRs) into a route table and associate the table with subnets.

Route tables act on the subnets they're associated with

You create a route table, add routes to it and associate it with zero or more subnets; each subnet can have zero or one route table (Virtual network traffic routing[26]). That count is of route table resources you create to hold UDRs, not of the system routes from the section above: Azure assigns system routes to every subnet automatically, so a subnet with zero route tables still routes by them (Virtual network traffic routing[26]). The route table must exist in the same subscription and region as the virtual network (Add, change, or delete a virtual network subnet[4]). Azure routes outbound traffic from a subnet based on the routes in that subnet's route table (Virtual network traffic routing[26]), so associate the table with every source subnet whose outbound path should change. The table's routes are combined with the subnet's default routes rather than replacing them, and where they conflict, the UDRs override the defaults (Virtual network traffic routing[26]). A route table holds up to 400 UDRs by default, or up to 1,000 through an Azure Virtual Network Manager routing configuration (Virtual network traffic routing[26]).

Next hop types you can choose

The next hop types available in a UDR are these (Virtual network traffic routing[26]):

Next hop type (CLI and PowerShell name) Sends the traffic to
Virtual appliance (VirtualAppliance) A private IP address you specify: an NVA's NIC or an internal load balancer
Virtual network gateway (VirtualNetworkGateway) The virtual network's gateway, supported only when that gateway is a VPN gateway
Virtual network (VNetLocal) Routing within the virtual network, overriding a broader route
Internet (Internet) The internet, explicitly
None (None) Nowhere: the traffic is dropped

Two hop types are absent on purpose: you can't specify Virtual network peering or VirtualNetworkServiceEndpoint in a UDR, because Azure creates those routes only when you configure a peering or a service endpoint (Virtual network traffic routing[26]). The gateway restriction matters as well: a UDR with next hop Virtual network gateway isn't supported when the network's gateway is an ExpressRoute gateway, a Route Server or a Virtual WAN hub router (Virtual network traffic routing[26]), so you can't force traffic to an ExpressRoute gateway with a UDR.

Sending everything to one device, and the exceptions

A UDR for 0.0.0.0/0 with next hop Virtual appliance or Virtual network gateway (often called forced tunneling) sends every destination that lacks a more specific route to that device, including the public IP addresses of Azure services not reached through a service endpoint, and resources in the subnet can no longer be reached directly from the internet (Virtual network traffic routing[26]). Longer prefixes carve exceptions out of it: a route with next hop Internet explicitly sends a named prefix to the internet, and a route with next hop None discards a prefix instead of forwarding it (Virtual network traffic routing[26]). Permission doesn't change paths: an NSG allow rule never alters the next hop a route selected.

Service tags as route prefixes

A UDR's address prefix can be a service tag instead of an explicit range; Microsoft manages the prefixes behind the tag and updates them as addresses change, and each route table supports 25 or fewer service-tag routes (Virtual network traffic routing[26]). On an exact prefix match, a route with an explicit IP prefix beats a service-tag route; among matching service-tag routes the order is regional tags (such as Storage.EastUS), then top-level tags (Storage), then regional AzureCloud tags, then AzureCloud (Virtual network traffic routing[26]). Choose the narrowest tag that covers the traffic: AzureCloud represents all Azure datacenter public IP addresses (Virtual network service tags[28]), so a route meant only for Storage traffic should use Storage.

Listing: a UDR in the route table MyRouteTable that sends traffic for the Storage service tag to an appliance at 10.0.100.4 (Azure CLI)

az network route-table route create \
  --resource-group MyResourceGroup \
  --route-table-name MyRouteTable \
  --name StorageRoute \
  --address-prefix Storage \
  --next-hop-type VirtualAppliance \
  --next-hop-ip-address 10.0.100.4

The service tag goes where an address prefix normally would, in --address-prefix; in Azure PowerShell, New-AzRouteConfig takes it the same way through -AddressPrefix (Virtual network traffic routing[26]).

Gateway route propagation

A route table also controls whether its subnets learn routes from VPN Gateway and ExpressRoute: disabling virtual network gateway route propagation on the table stops Azure adding those routes, both static and BGP, to the associated subnets (Virtual network traffic routing[26]). Never disable it on GatewaySubnet, or the gateway doesn't function (Virtual network traffic routing[26]); a UDR with a 0.0.0.0/0 destination on the gateway subnet isn't supported either (About VPN Gateway configuration settings[6]).

A UDR changes a path only for the subnets whose route table holds it, only with the next hop types listed above, and only where it wins the longest-prefix contest.

Routing through a network virtual appliance

Sending traffic through a firewall VM is one use of UDRs, and it works only when the Azure route, the Azure NIC setting and the appliance's own software all agree.

An NVA is a VM that runs a network application, such as a firewall (Virtual network traffic routing[26]). The route to it uses next hop type Virtual appliance plus a next hop IP address: the private IP of the NVA's NIC, or the private IP of an internal load balancer in front of several NVAs; a route for 0.0.0.0/0 with this next hop lets the appliance inspect all such traffic and decide whether to forward or drop it (Virtual network traffic routing[26]). Four requirements follow:

  1. Deploy the NVA in a different subnet from the resources that route through it. Putting it in the same subnet and applying the route table there can cause routing loops in which traffic never leaves the subnet (Virtual network traffic routing[26]).
  2. Make the next hop IP directly reachable, not through an ExpressRoute gateway or Azure Virtual WAN; otherwise the UDR configuration is invalid (Virtual network traffic routing[26]).
  3. Enable IP forwarding on the NVA's Azure NIC. This Azure setting disables the platform's source and destination check for the NIC (Virtual network traffic routing[26]), so it can receive traffic not addressed to it and send traffic with a different source address; enable it on every NIC that forwards traffic (Create, change, or delete a network interface[9]).
  4. Configure forwarding inside the VM as needed. The Azure setting doesn't configure the guest: you might need to enable IP forwarding in the operating system (Virtual network traffic routing[26]), and the VM must run an application able to forward the traffic (Create, change, or delete a network interface[9]).

When the NVA sits in a peered hub, the peering adds its own pieces: a UDR's next hop can be the IP address of a VM in a peered network (Azure virtual network peering[21]), and the link of each network that receives the forwarded traffic needs Allow forwarded traffic (see the peering settings section above). Enabling Allow forwarded traffic alone does nothing for an appliance whose NIC and operating system don't forward. Because a route table acts only on traffic leaving the subnets it is associated with, the return direction may need routes on the destination subnets too if the existing return path would bypass the appliance, so that a stateful appliance (one that tracks each connection) sees both directions of it.

The figure below follows one packet from a workload subnet through the appliance.

If traffic still bypasses the appliance, look in the effective routes for a more specific route, such as a peering route, that wins over your UDR (see the troubleshooting section below).

In short, all the pieces must agree: a UDR in each source subnet, IP forwarding on the NVA's Azure NIC and in its operating system, and, across a peering, Allow forwarded traffic plus any return routes the path needs.

VM in a workload subnetRoute table: 0.0.0.0/0next hop VirtualApplianceNVA NIC: IP forwardingenabled in AzureNVA operating systemforwards the packetDestinationNVA in its own subnet,separate from the workloadsthat route through it
One packet routed through an NVA: the UDR in the workload subnet, IP forwarding on the NVA's Azure NIC, then forwarding in the NVA's operating system

Outbound internet access: private subnets and NAT gateway

A VM that calls an internet API or downloads updates needs a public source address, and how it gets one depends on the subnet's outbound configuration.

Default outbound access and private subnets

When a VM is deployed in a nonprivate subnet without an explicit outbound connectivity method, Azure can assign it a default outbound public IP address that is owned by Microsoft and can change without notice; this is default outbound access (Default outbound access in Azure[29]). The explicit methods include (Default outbound access in Azure[29]):

  • a NAT gateway associated with the subnet;
  • a public IP address associated with the VM's network interface;
  • outbound rules on a Standard public load balancer whose backend pool contains the VM.

A private subnet is one whose defaultOutboundAccess property is false; private refers to outbound access, not to the subnet's address range. Its VMs can't use default outbound access, so general public-endpoint connectivity needs an explicit method, although same-region Azure Storage remains reachable without one, and some services, such as Windows Activation and Windows Update, don't function without explicit egress (Default outbound access in Azure[29]). The Azure portal already defaults new subnets to private, and with the virtual network API version released after March 31, 2026, subnets in new virtual networks are private by default; deployments that use earlier API versions keep the earlier behavior, and existing virtual networks aren't changed (Default outbound access in Azure[29]). To remove default outbound IPs completely from an existing subnet, the subnet must be made private and its VMs stopped and deallocated (Default outbound access in Azure[29]), so configure the explicit method before that transition.

An NSG rule that allows outbound internet traffic doesn't fix a private subnet: it permits traffic but provides no public address translation (source network address translation, SNAT).

NAT gateway

A NAT gateway attached to a subnet lets all instances in that subnet connect outbound to the internet while staying private, and it doesn't permit unsolicited inbound connections from the internet: only response packets to an outbound connection pass through (What is Azure NAT Gateway?[30]). NAT gateway has two SKUs, Standard (zonal) and Standard v2 (zone redundant) (What is Azure NAT Gateway?[30]). Eligible subnets of a virtual network, excluding GatewaySubnet and subnets containing SQL managed instances, can use the same NAT gateway, and a Standard NAT gateway supports up to 16 IPv4 public IP addresses, individually or from a public IP prefix (What is Azure NAT Gateway?[30]). It is the usual answer for scalable, outbound-only access through controlled source addresses, instead of a public IP per VM. Inbound connections need something else, such as a public IP on the VM's NIC or a public load balancer, plus NSG rules that allow the traffic. An inbound NAT rule is not an outbound method: it forwards a load balancer frontend port to a backend VM (Inbound NAT rules[31]); see 'Inbound NAT rule' in Configure name resolution and load balancing.

Which method wins

When several methods apply, the precedence is (What is Azure NAT Gateway?[30]): a UDR that sends 0.0.0.0/0 to a virtual appliance or a virtual network gateway, then the NAT gateway, then an instance-level public IP on the VM, then load balancer outbound rules, then the default system route to the internet (which, with no other method present, is default outbound access). Adding a NAT gateway therefore changes the egress address for new outbound connections even for VMs that also have public IPs, and a UDR that sends 0.0.0.0/0 to an appliance determines the path ahead of the NAT gateway. The figure below shows the order; its last step is unavailable in a private subnet.

Decide egress explicitly: in a private subnet internet egress needs an explicit method, such as the three listed above or a UDR through a firewall or NVA, and when methods overlap, the precedence order decides.

1. UDR to an appliance ora virtual network gatewayif none2. NAT gatewayif none3. Public IP on the NICif none4. Load balanceroutbound rulesif none5. Default outbound access(not in a private subnet)
Outbound connectivity precedence for a VM's internet-bound traffic, from a forcing UDR down to default outbound access

Troubleshooting connectivity layer by layer

When a VM can't reach a destination, check addresses, then routes, then rules, then live tests and evidence, because the tools for routes, rules and live tests answer different questions: Next hop checks whether traffic is routed correctly to the intended destination, IP flow verify checks whether a packet is allowed or denied, and Connection troubleshoot tests a connection at a point in time (What is Azure Network Watcher?[32]). Addresses come first because the later checks start from one: the NIC's IP configuration holds a private address from the address space of the subnet the NIC is deployed in (Configure IP addresses for an Azure network interface[33]), Next hop asks for the VM's source IP address (Diagnose a virtual machine network routing problem[34]), and IP flow verify tests rules using the local and remote IP addresses, ports, protocol and traffic direction (IP flow verify overview[35]).

The main Azure Network Watcher tools used here, which the figure below places beside the layer each one checks, monitor and diagnose IaaS resources such as VMs and virtual networks, and Network Watcher is enabled automatically in a region when you create or update a virtual network there (What is Azure Network Watcher?[32]). Connection monitor, unlike Connection troubleshoot, monitors a connection over time, and flow logs store IP traffic data in Azure Storage, which traffic analytics visualizes (What is Azure Network Watcher?[32]); see Monitor resources in Azure.

1. Addresses

Confirm that the VM's NIC is in the expected subnet with the expected IP configuration, and that any address set inside the guest matches the one Azure assigned (see the section on private IP address allocation above).

2. Routes

Two views answer different questions. A NIC's Effective routes page lists the routes in force for the NIC: a combination of the Azure default routes, any user-defined routes and any BGP routes for the NIC's subnet, shown when the NIC is attached to a running VM (Create, change, or delete a network interface[9]). A route table resource lists only the UDRs you put in it, so it can't show whether a system or peering route wins. Network Watcher Next hop answers for one destination: it returns the next hop type, IP address and route table ID for a specific destination IP address (What is Azure Network Watcher?[32]); when a UDR decides, it names the route table holding that route, and otherwise it reports System Route (Next hop overview[36]). For peering problems, by default a peered network should appear in the effective routes with next hop type Virtual network peering; if peering route propagation is disabled, check the replacement UDRs instead (Azure virtual network peering[21]).

3. Rules

A correct route doesn't mean the traffic is allowed through an NSG, an appliance or the guest. IP flow verify checks whether a packet to or from a VM is allowed or denied and tells you which security rule decided (What is Azure Network Watcher?[32]); it evaluates security rules, not routes, so it never identifies a next hop. Network Watcher's Effective security rules tool shows all security rules applied to a network interface, to the subnet the interface is in, and the aggregate of both (What is Azure Network Watcher?[32]); see Configure secure access to virtual networks.

4. Live tests and evidence

The remaining tools test or record actual traffic, or visualize network configuration (What is Azure Network Watcher?[32]):

Tool Use it when
Connection troubleshoot You need a point-in-time connection test from a VM, scale set, Application Gateway v2 (except Private Application Gateway deployments) or Bastion host to a VM, FQDN, URI or IPv4 address
Connection monitor You need continuous end-to-end connection monitoring, including network performance, for Azure and hybrid endpoints
Packet capture You need the actual packets to and from a VM or scale set
Flow logs You need a log of IP traffic flows, not packet contents
Topology You need a visualization of resources and their relationships, not a test

Choose by the question: immediate diagnosis or continuing evidence, flow records or packets. Packet capture has prerequisites: it requires the Network Watcher agent VM extension AzureNetworkWatcherExtension, can filter on protocol, local and remote IP address and port, and saves captures to the VM's local disk or to Azure Blob Storage (Packet capture overview[37]). It reaches the storage account with shared access signature (SAS) tokens, so key access must be enabled on the account; without it, captures can only be saved to the local disk (Packet capture overview[37]). For flow records, use virtual network flow logs: NSG flow logs retire on September 30, 2027 and no longer support creating new ones (What is Azure Network Watcher?[32]).

Resolve each failure before moving on from addresses to routes, rules, and then live tests and evidence: a correct next hop says nothing about security rules, and IP flow verify tests the rules of every network security group that applies to the VM's network interface, whether that group is associated with the VM's subnet or with the network interface itself (IP flow verify overview[35]). Connection troubleshoot reaches further: it also detects virtual machine (guest) firewall rules blocking traffic and servers not listening on designated destination ports (Connection troubleshoot overview[38]).

1. AddressesNIC IP configuration2. RoutesEffective routes, Next hop3. RulesIP flow verify,Effective security rules4. Live tests and evidenceConnection troubleshoot,Connection monitor,Packet capture
The four troubleshooting layers for VM connectivity and the Network Watcher or NIC tool that checks each one

Connecting networks and clients: system routes vs virtual network peering vs site-to-site VPN vs Azure Bastion vs point-to-site VPN

QuestionSystem routesVirtual network peeringSite-to-site VPNAzure BastionPoint-to-site VPN
What it connectsSubnets of one virtual networkTwo Azure virtual networks, in the same or different regionsAn on-premises site's network and a virtual networkAn administrator and VMs, for RDP or SSH sessions onlyOne remote computer and a virtual network
What you deployNothing; Azure creates the routesA peering link on each of the two virtual networksA VPN gateway plus an on-premises VPN deviceAzure Bastion, deployed in the virtual networkA VPN gateway plus a VPN client on the computer
Path the traffic takesInside the virtual networkMicrosoft backbone, private addresses, no gatewayIPsec/IKE tunnel over the public internetRDP or SSH over TLS, then to the VMs' private IP addressesOpenVPN, IKEv2 or SSTP tunnel over the public internet
Typical requirementAn app tier calls a database tier in another subnetWorkloads in two virtual networks talk privatelyA branch office reaches Azure workloadsAdministrators manage VMs that have no public IPA remote employee reaches private applications

Decision tree

Both ends in onevirtual network?YesSystem routesNoOther end an Azurevirtual network?YesVirtualnetworkpeeringNoWhole on-premisessite, VPN device?YesSite-to-siteVPNNoOnly RDP or SSHto VMs needed?YesAzure Bastion(VM sessions)NoPoint-to-site VPN(one computer)Spoke to spoke via a hub:peering plus NVA and UDRs

Sharp facts the exam loves — give these one last read before exam day.

Cheat sheet

Sharp facts the exam loves — scan these before test day.

A virtual network is regional and its subnet address ranges must not overlap

An Azure virtual network belongs to one region but spans that region's availability zones. Each subnet must fit within the virtual network address space and must not overlap another subnet in that network. Using a publicly routable address range inside a virtual network does not make its resources directly reachable from the internet; private addressing is recommended for normal deployments.

2 questions test this
Subnet sizing must allow for Azure's five reserved IPv4 addresses

Azure reserves the first four IPv4 addresses and the last IPv4 address in each subnet. The smallest supported IPv4 subnet is /29, while IPv6 subnets must be /64. Size a subnet for usable addresses after the reservations and for any service-specific minimum, rather than treating every address in the prefix as assignable.

2 questions test this
Multiple subnet prefixes can expand eligible VM subnets without removing existing resources

For an eligible VM or VM scale-set subnet, add address prefixes to increase capacity without first clearing its existing resources. Update addressPrefixes through CLI, PowerShell, or an ARM template and include every prefix to retain; omitted unused prefixes are removed, while omitting a prefix used by a NIC produces an error. This feature restricts each NIC to one IPv4 and one IPv6 address and excludes delegated subnets except GatewaySubnets delegated to ExpressRoute Gateway services. Replacing an in-use range outside this capability still requires moving or removing affected resources, and deleting a subnet requires clearing its resources.

2 questions test this
A VM connects directly to different subnets through separate network interfaces

When a VM must connect directly to two subnets, attach a network interface for each subnet. All interfaces attached to that VM must belong to the same virtual network, and the VM size must support the required interface count. Stop and deallocate an existing VM before adding or removing interfaces, then configure its guest operating system to use the additional interfaces.

Trap Creating a network bridge in the guest operating system, which does not attach an Azure network interface to the second subnet.

2 questions test this
Additional IP configurations add addresses to one NIC rather than connect it to another subnet

An Azure network interface has one primary IP configuration and can have secondary IP configurations where the subnet's capabilities permit multiple addresses. Its private IP configurations use the same subnet, so additional configurations serve multiple addresses on that subnet rather than multiple subnet attachments. Add the secondary private addresses to the guest operating system as required, but never configure the Azure public IP addresses in the guest.

2 questions test this
Static private addresses are selected in the Azure NIC configuration

Azure dynamic private allocation selects an available address from the subnet; static private allocation lets you select an unused, nonreserved address in that subnet. Set the allocation on the network interface rather than changing only the guest operating system. A guest address that differs from the Azure-assigned address can make the VM unreachable, so any necessary guest configuration must match Azure's configuration.

2 questions test this
Subnet delegation authorizes a service to deploy its service-specific resources

Subnet delegation gives an Azure service explicit permission to create service-specific resources in a selected subnet. Configure the delegation required by the service rather than treating a service endpoint as permission to deploy into the subnet. If service resources already occupy a delegated subnet, remove those resources before changing the delegation.

Trap Enabling a service endpoint for subnet injection, which enables access to a service endpoint instead of delegating the subnet for that service's deployments.

2 questions test this
Az PowerShell subnet configuration changes must be saved to the virtual network

Add-AzVirtualNetworkSubnetConfig adds a subnet configuration to an in-memory virtual network object; Set-AzVirtualNetwork saves that object's changes to Azure. The same retrieve, modify, and save pattern applies when removing a subnet configuration. Azure CLI uses az network vnet subnet create or delete against the named virtual network directly, so the command's target must identify both the resource group and virtual network.

2 questions test this
A dynamic private address remains allocated to a retained NIC through VM deallocation

A NIC's dynamically assigned private IP address is not released merely because its VM is stopped and deallocated. Deleting the NIC, moving it to another subnet in the same virtual network, or switching to static allocation with a different specified address releases or changes the old assignment. Switching from dynamic to static without specifying a different address retains the current address by default, so changing allocation mode need not renumber the VM.

2 questions test this
Accelerated Networking reduces host processing without removing network policy enforcement

When a supported VM needs lower network latency, less jitter, and lower CPU overhead for packet processing, enable Accelerated Networking on its NIC. The accelerated data path bypasses the host virtual switch while enforcing its network policies in hardware, so NSGs and isolation still apply. Confirm that the VM size and guest operating system support the feature, and stop and deallocate an existing VM before enabling it; a running VM cannot be enabled directly.

2 questions test this
A VM, its NICs, and their virtual network must share a subscription and region

A VM's network interfaces must be in the same subscription and Azure region as the VM, and each interface's virtual network must share that subscription and region. The NIC, VM, and virtual network can belong to different resource groups. An existing NIC can change subnets within its virtual network, but its virtual network assignment cannot be changed as an in-place subnet update.

Trap Using global VNet peering to make a remote-region NIC eligible for direct VM attachment, although peering provides routed connectivity rather than changing NIC placement requirements.

2 questions test this
IPv6 VM connectivity requires an IPv4 configuration on every NIC

Azure IPv6 VM networking is dual-stack: every NIC must retain at least one IPv4 IP configuration, and IPv6-only VMs and VM scale sets are unsupported. Add IPv6 through a secondary NIC IP configuration only when the virtual network and subnet also have IPv6 address space.

2 questions test this
Peering connects nonoverlapping virtual networks over the Microsoft backbone

Virtual network peering requires nonoverlapping address spaces. Regional peering connects networks in the same region, and global peering connects networks in different regions; supported configurations can cross subscriptions and Microsoft Entra tenants. Traffic uses private addresses over the Microsoft backbone without requiring public IP addresses or a gateway for the peering itself.

Trap Using a private endpoint to connect entire virtual network address spaces, although it exposes a selected service rather than providing general network peering.

2 questions test this

A peering connection consists of a peering resource in each virtual network. Creating only one side leaves the relationship in Initiated rather than Connected; create the reciprocal side and verify Connected status. Add-AzVirtualNetworkPeering and az network vnet peering create create a directional link, so command-based deployment must configure both networks.

2 questions test this
Peering does not automatically provide transit between a hub's peers

Peering VNet1 with VNet2 and VNet2 with VNet3 does not directly connect VNet1 with VNet3. Create an explicit VNet1-to-VNet3 peering for direct connectivity, or configure routing through a network virtual appliance when inspected transit is required. The hub's existing peering relationships alone do not supply the transit path.

Trap Enabling gateway transit for ordinary spoke-to-spoke forwarding through a hub VM, although gateway transit shares a gateway rather than making peering transitive.

2 questions test this
Forwarded-traffic permission and transit routing are separate peering requirements

When traffic reaches a peer after being forwarded by an appliance in another network, enable the receiving peering's option to allow forwarded traffic. This setting is separate from allowing access between the peer networks and does not create an appliance or user-defined route. Configure the forwarding appliance and routes independently, including the return path.

Trap Selecting Use remote gateways for traffic forwarded by a firewall VM, which selects shared gateway use rather than admitting appliance-forwarded traffic.

3 questions test this
Gateway transit pairs the gateway owner's permission with the peer's remote-gateway setting

To share a hub gateway, enable Allow gateway transit on the hub's peering and Use remote gateways on the consuming network's peering. A consuming network cannot have its own gateway and can select a remote gateway on only one peering. Gateway transit is supported with regional and global peering, allowing the peer to reach networks connected through the shared gateway.

Trap Enabling Allow forwarded traffic alone for shared VPN connectivity, which does not select the hub gateway or establish gateway transit.

2 questions test this
Peers must synchronize after a peered virtual network's address space changes

You can resize a Resource Manager virtual network's address space while it is peered without interrupting connectivity to its existing address space. Synchronize the peers after each address-space change so they learn the updated ranges. Check the synchronization state and effective routes when workloads in a newly added range cannot be reached across an otherwise Connected peering.

2 questions test this
Cross-subscription peering requires authorization to configure both sides

Peering can cross subscription and tenant boundaries, but the administrators must have the required peering permissions on the respective virtual networks. Network Contributor is a suitable built-in role; a known remote resource ID can identify a network that is not visible in the portal selector, but does not grant permission to modify it. Coordinate creation of the reciprocal link with the remote administrator when one account does not control both networks.

Trap Moving both virtual networks into one resource group to enable peering, although authorization and nonoverlapping addresses, not common resource-group membership, determine eligibility.

2 questions test this
Standard public IP resources retain a static address until the IP resource is deleted

Standard public IPv4 and IPv6 addresses use static allocation, with Azure selecting the address when the public IP resource is created. Stopping a VM does not release the address of its retained Standard public IP resource; deleting an individually allocated IP resource releases its address and it cannot be recovered. An address allocated from a retained public IP prefix follows that prefix's separate reservation lifecycle. Basic public IPs are retired, so new VM and load-balancer configurations should use the supported Standard resources rather than a Basic dynamic address.

2 questions test this
A VM public IP is associated with a NIC IP configuration and still requires an NSG allow rule

For direct internet connectivity to a VM, associate a public IP resource with the required network interface IP configuration. Standard public IP addresses are closed to inbound traffic by default, so an associated subnet or NIC NSG must explicitly allow the required inbound traffic. Configure the guest firewall and listener as well; associating the public IP supplies addressability rather than opening an application port.

1 question tests this
Zone-redundant public IPs protect the address frontend against a single-zone failure

In regions with availability-zone support, a Standard public IP can be zonal or zone-redundant. A zonal IP shares the fate of its selected zone, whereas a zone-redundant IP survives a single-zone failure; formerly non-zonal Standard IPs in supported regions are now zone-redundant by default. The IP's zone resilience does not make a single-zone backend resilient, and a public IP resource's zone selection cannot be changed after creation.

Trap Relying on a VM availability set to make a zonal public IP zone-redundant, although backend placement does not alter the IP resource's zone configuration.

2 questions test this
A public IP DNS label creates an Azure-managed regional hostname

A public IP DNS label creates a hostname in the form label.region.cloudapp.azure.com, and the label must be unique within its Azure region. Azure creates an A or AAAA mapping according to the IP version. Use records in your own DNS zone when clients require your organization's domain rather than the Azure-managed hostname.

Trap Creating a private DNS virtual network link to publish a public IP's custom internet name, although such a link exposes private-zone resolution to a virtual network rather than publishing public records.

2 questions test this
A public IP prefix reserves a contiguous address range for stable allowlists

When external firewalls must allow a predictable range as resources scale, allocate Standard public IP addresses from a public IP prefix. Normal IPv4 prefix sizes are /28, /29, /30, and /31, and the prefix size cannot be changed after creation. Deleting an individual public IP returns its address to the retained prefix's reserved range, allowing reuse while the prefix remains allocated.

Trap Using a service tag to reserve contiguous public addresses for your VMs, although a service tag represents Microsoft-maintained service prefixes rather than a customer allocation.

2 questions test this
Internet routing preference trades Microsoft backbone carriage for lower-cost ISP carriage

A public IP's default Microsoft network routing preference keeps traffic on Microsoft's network for most of the path, while Internet preference hands traffic to the ISP network closer to the Azure region for cost-optimized egress. Select routing preference at creation because it cannot be changed on the existing IP. Internet preference supports only eligible Standard zone-redundant IPv4 resources, not IPv6 or NAT Gateway, and traffic destined within Azure still takes the direct Microsoft network path.

Trap Changing a subnet's user-defined default route to Internet to select lower-priced ISP routing, although that next-hop setting is not the public IP's routing-preference property.

2 questions test this
Public IP tier and SKU must match the consuming network service

Use a Regional Standard public IP for a regional Standard public Load Balancer; the Global tier is intended for a cross-region load balancer frontend. Standard v2 public IPs are restricted to Standard v2 NAT Gateway rather than serving as a general replacement for Standard IPs on VMs or load balancers. Select the IP resource for the target service's supported tier and SKU rather than choosing solely by the word Standard.

Trap Selecting a Global-tier IP to make a regional load balancer automatically span regions, although the cross-region load-balancer configuration is a separate requirement.

2 questions test this
Public IP moves preserve regional identity but require compatible association state

A regional public IP must be in the consuming resource's subscription and region, although it can be in another resource group. A Standard public IP can move across resource groups while attached except when attached to a virtual network gateway, but it must be detached before moving across subscriptions and cannot move across regions. Dissociate an address from its IP configurations before deleting the public IP resource, so a reorganization that must preserve the address should use a supported move rather than deletion and recreation.

2 questions test this
An existing Basic public IP requires the migration path supported by its consuming resource

When upgrading an existing Basic public IP, identify the consuming resource and schedule downtime before choosing the migration procedure. A standalone public IP resource must use static allocation and be disassociated before its one-way upgrade to Standard, which preserves the allocated address; set static allocation before detaching when address identity must be retained. A Basic Load Balancer requires migration to a Standard Load Balancer, while a gateway requires its gateway-specific migration rather than an independent attached-IP upgrade. Prepare the required NSG allow rules for the Standard frontend before restoring traffic, because Standard ingress is secure by default.

Trap Using deletion and recreation to preserve an individually allocated public address, which releases the allocation instead of upgrading the retained resource.

User-defined routes take effect through a subnet's associated route table

Create a route table, add user-defined routes, and associate it with each source subnet whose outbound routing must change. A subnet can have at most one associated route table, but a table can serve multiple subnets in the same subscription and region. Custom routes combine with Azure's system routes rather than replacing the entire routing table. A route table normally supports 400 UDRs, with support for up to 1,000 through Azure Virtual Network Manager routing configuration.

2 questions test this
System routes already provide intra-network routing and discard unconnected private ranges

Azure automatically routes between subnets within a virtual network using Virtual network system routes, so ordinary inter-subnet connectivity does not require a gateway. System routes with next hop None discard traffic to reserved private ranges that are not otherwise reachable through a more appropriate route. The default Internet route handles other destinations, but a route alone does not supply the explicit outbound address translation needed by a private subnet.

Trap Adding a VPN gateway for normal traffic between subnets of one virtual network, although system routes already provide that Layer 3 path.

2 questions test this
Azure normally selects the longest matching route before considering route source

Azure normally selects the route with the longest matching destination prefix; a broad 0.0.0.0/0 user-defined route does not override a more-specific peering route. For identical prefixes, the normal preference is user-defined route, then BGP route, then system route. Azure's preferred system routes for virtual networks and peering take precedence over BGP routes even when BGP is more specific; service endpoint routes are also preferred and cannot be overridden.

2 questions test this
Routing through an appliance requires forwarding on both Azure and the appliance

To send traffic through a network virtual appliance, use next hop VirtualAppliance and the appliance's private IP address, then enable IP forwarding on the appliance's Azure NIC. Configure forwarding in its operating system or network software as required; the Azure NIC setting does not configure that software. Keep the appliance in a separate subnet from the workloads whose route table points to it, and ensure the next-hop address is directly reachable to avoid loops or an invalid route.

Trap Enabling only the peering's forwarded-traffic option for a firewall VM, which does not enable packet forwarding on the appliance's NIC or operating system.

2 questions test this
A UDR can target a VPN gateway but cannot manually create a peering or service-endpoint hop

The configurable UDR next-hop types are VirtualAppliance, VirtualNetworkGateway, VNetLocal, Internet, and None. A VirtualNetworkGateway UDR is supported for a VPN gateway, not for selecting an ExpressRoute gateway, Route Server, or Virtual WAN hub router. Virtual network peering and VirtualNetworkServiceEndpoint hops are generated by their respective Azure features rather than selected as UDR next-hop types.

Trap Selecting VirtualNetworkGateway to force traffic to an ExpressRoute gateway, although that UDR next-hop mechanism supports VPN gateways instead.

2 questions test this
None routes discard a prefix while Internet routes create direct-routing exceptions

Use next hop None when a matching destination prefix must be discarded, or Internet when an explicit destination should take the direct Internet path instead of a broader forced-tunneling route. A 0.0.0.0/0 route to an appliance changes the path for destinations not matched by a more-specific route. Ensure return traffic follows a path compatible with stateful inspection, because routing only one direction through a stateful device can break the connection.

Trap Using an NSG allow rule to bypass a forced-tunneling route for an allowed destination, although permission to send traffic does not change its selected next hop.

2 questions test this
Gateway route propagation is controlled per subnet through its route table

Disable virtual network gateway route propagation on a subnet's route table when that subnet must not automatically learn VPN or ExpressRoute gateway routes. The setting affects propagated static and BGP gateway routes, not just one advertised prefix. Do not disable propagation on GatewaySubnet or assign it a 0.0.0.0/0 UDR, because these configurations can prevent a VPN gateway from functioning.

2 questions test this
Service-tag UDRs let Microsoft maintain the destination service prefixes

A service tag can replace an explicit address prefix in a UDR, allowing Microsoft to update the underlying Azure service address ranges. For the same matching prefix, an explicit IP-prefix route takes precedence over a service-tag route; among matching tags, regional service tags precede service-wide tags, then regional AzureCloud, then AzureCloud. A route table supports up to 25 service-tag routes, and New-AzRouteConfig or az network route-table route create supplies the service tag as the address-prefix value.

Trap Using AzureCloud instead of Storage for a storage-only appliance route, although AzureCloud also includes non-Storage service prefixes.

2 questions test this
Effective routes explain a NIC's route set while Next hop identifies the selected destination path

When troubleshooting routing, view the NIC's effective routes to see system, user-defined, and gateway-propagated routes together; the custom route-table resource lists only its configured UDRs. Use Network Watcher Next hop for a particular destination to identify the selected next-hop type, address, and route-table ID. An expected route does not establish that an NSG, appliance, or guest listener will permit the traffic.

Trap Using IP flow verify to identify a routing next hop, although that tool evaluates security-rule permission rather than route selection.

3 questions test this
Connection troubleshoot tests now while Connection monitor tracks reachability over time

Use Network Watcher Connection troubleshoot for a point-in-time connectivity test from a supported Azure source to a target such as a VM, FQDN, or IP address. Use Connection monitor when the requirement is ongoing endpoint reachability and network-performance monitoring, including hybrid endpoints. Select the tool by whether you need immediate diagnosis or continuing evidence rather than substituting a configuration-only route or rule view.

Trap Using the Topology visualization for latency monitoring, although it shows resource relationships rather than repeatedly testing an end-to-end connection.

3 questions test this
Network Watcher packet capture provides packet evidence rather than only flow summaries

Use Network Watcher packet capture when diagnosis requires the actual packets exchanged by a VM or scale-set instance. It requires AzureNetworkWatcherExtension, supports filters for protocol, addresses, and ports, and can save captures locally or to Blob Storage. Storage-account key access must be enabled for the SAS-based storage destination; otherwise use the VM's local disk, while virtual network flow logs remain the appropriate source for broad historical flow metadata.

Trap Using virtual network flow logs to inspect application packet contents, although flow metadata does not contain the complete packet capture.

3 questions test this
Private subnets require an explicit outbound method even when security rules allow internet traffic

A subnet with defaultOutboundAccess set to false does not provide implicit outbound connectivity; configure NAT Gateway, a suitable public IP, or Standard public Load Balancer outbound rules when its VMs need public endpoints. The portal defaults new subnets to private, and the virtual network API released after March 31, 2026 uses that default; older API versions and existing networks retain their earlier behavior. Making an existing subnet private and stopping and deallocating its VMs clears retained default outbound assignments, so establish explicit egress before the transition.

Trap Adding an outbound Internet allow rule to restore a private subnet's missing egress, although an NSG rule does not create a public translation method.

3 questions test this
NAT Gateway supplies subnet-wide outbound access without exposing inbound VM ports

For scalable outbound-only connectivity with controlled public source addresses, associate NAT Gateway with the subnet rather than assigning each VM a public IP. For internet-bound new flows on the direct Internet path, NAT Gateway takes precedence over instance public IPs and Load Balancer outbound SNAT; a UDR that sends traffic to another next hop still determines the routing path. NAT Gateway does not accept unsolicited inbound connections, while a public Load Balancer or a VM public IP can support separately authorized inbound traffic.

Trap Creating inbound NAT rules to supply general outbound access, although those rules forward selected inbound ports and do not define outbound SNAT.

4 questions test this
VPN Gateway connects remote clients and on-premises sites through different connection types

For encrypted network connectivity from an on-premises site over the public internet, use a site-to-site VPN between Azure VPN Gateway and the site's VPN device. For an individual remote client that needs access to the virtual network, use point-to-site VPN with a supported client protocol rather than requiring a site VPN device. Bastion is appropriate for remote VM management instead of general client network access, and a public IP alone supplies neither an encrypted network tunnel nor private-network routes.

Trap Using Azure Bastion as a general remote-user VPN for private applications, although Bastion provides VM management sessions rather than broad client-to-network connectivity.

3 questions test this

References

  1. https://learn.microsoft.com/en-us/credentials/certifications/resources/study-guides/az-104
  2. https://learn.microsoft.com/en-us/azure/virtual-network/virtual-networks-faq
  3. https://learn.microsoft.com/en-us/azure/vpn-gateway/nat-overview
  4. https://learn.microsoft.com/en-us/azure/virtual-network/virtual-network-manage-subnet
  5. https://learn.microsoft.com/en-us/azure/virtual-network/how-to-multiple-prefixes-subnet
  6. https://learn.microsoft.com/en-us/azure/vpn-gateway/vpn-gateway-about-vpn-gateway-settings
  7. https://learn.microsoft.com/en-us/azure/expressroute/expressroute-introduction
  8. https://learn.microsoft.com/en-us/powershell/module/az.network/add-azvirtualnetworksubnetconfig
  9. https://learn.microsoft.com/en-us/azure/virtual-network/virtual-network-network-interface
  10. https://learn.microsoft.com/en-us/azure/virtual-network/virtual-network-network-interface-vm
  11. https://learn.microsoft.com/en-us/azure/virtual-network/ip-services/virtual-network-multiple-ip-addresses-portal
  12. https://learn.microsoft.com/en-us/azure/virtual-network/ip-services/ipv6-overview
  13. https://learn.microsoft.com/en-us/azure/virtual-network/accelerated-networking-overview
  14. https://learn.microsoft.com/en-us/azure/virtual-network/ip-services/private-ip-addresses
  15. https://learn.microsoft.com/en-us/azure/virtual-network/ip-services/public-ip-addresses
  16. https://learn.microsoft.com/en-us/azure/virtual-network/ip-services/virtual-network-public-ip-address
  17. https://learn.microsoft.com/en-us/azure/virtual-network/ip-services/public-ip-basic-upgrade-guidance
  18. https://learn.microsoft.com/en-us/azure/virtual-network/ip-services/routing-preference-overview
  19. https://learn.microsoft.com/en-us/azure/virtual-network/ip-services/public-ip-address-prefix
  20. https://learn.microsoft.com/en-us/azure/virtual-network/ip-services/public-ip-upgrade-portal
  21. https://learn.microsoft.com/en-us/azure/virtual-network/virtual-network-peering-overview
  22. https://learn.microsoft.com/en-us/azure/virtual-network/virtual-network-manage-peering
  23. https://learn.microsoft.com/en-us/cli/azure/network/vnet/peering
  24. https://learn.microsoft.com/en-us/azure/vpn-gateway/vpn-gateway-about-vpngateways
  25. https://learn.microsoft.com/en-us/azure/bastion/bastion-overview
  26. https://learn.microsoft.com/en-us/azure/virtual-network/virtual-networks-udr-overview
  27. https://learn.microsoft.com/en-us/azure/virtual-network/network-security-groups-overview
  28. https://learn.microsoft.com/en-us/azure/virtual-network/service-tags-overview
  29. https://learn.microsoft.com/en-us/azure/virtual-network/ip-services/default-outbound-access
  30. https://learn.microsoft.com/en-us/azure/nat-gateway/nat-overview
  31. https://learn.microsoft.com/en-us/azure/load-balancer/inbound-nat-rules
  32. https://learn.microsoft.com/en-us/azure/network-watcher/network-watcher-overview
  33. https://learn.microsoft.com/en-us/azure/virtual-network/ip-services/virtual-network-network-interface-addresses
  34. https://learn.microsoft.com/en-us/azure/network-watcher/diagnose-vm-network-routing-problem
  35. https://learn.microsoft.com/en-us/azure/network-watcher/ip-flow-verify-overview
  36. https://learn.microsoft.com/en-us/azure/network-watcher/next-hop-overview
  37. https://learn.microsoft.com/en-us/azure/network-watcher/packet-capture-overview
  38. https://learn.microsoft.com/en-us/azure/network-watcher/connection-troubleshoot-overview