SDWAN Tunnel on FortiGate firewall through azure load balancer

sidhrane 146 Reputation points
2026-11-15T20:39:40.0333333+00:00

Hi,

I am building SDWAN tunnel between on-prem and azure in below locations on FortiGate FWs in active-passive HA mode.

  1. Location 1: Site with express route (primary SDWAN path) and Internet (secondary SDWAN path)

Q1. Since SDWAN tunnel using express route will pass through azure internal load balancer. Can SDWAN VPN connection from on-prem to azure / azure to on-prem will be built on same front-end IP of azure internal load balancer (similar to external load balancer where same IP will be used for Inbound/outbound VPN connection)

  1. Location 2: Site with only Internet (primary and secondary SDWAN path)

In site with no express route, SDWAN tunnel will be used on internet. Like on-prem DC, we have two internet service provider with two different physically diversed path.

Q2. Similarly, do we have internet link in azure with different physical path?

Q3. Can I build two SDWAN Tunnel between on-prem and azure with two public front end IP on external load balancer in active passive mode.

on-prem FW1 > SDWAN Tunnel 1 > Azure public FIP 1 > Azure FortiGate FW1 (primary tunnel)

on-prem FW2> SDWAN Tunnel 2 > Azure public FIP 2 > Azure FortiGate FW1 (primary tunnel)

Q4. If SDWAN tunnel initiated from azure to on-prem, how azure external load balancer will use outbound internet IP from two public front end IPs for tunnel 1 or tunnel 2. ( Azure FGT FW1 > Azure public FIP 1 > on-prem FW1 and Azure FGT FW1 > Azure public FIP 2> on-prem FW1)

Azure Load Balancer
Azure Load Balancer

An Azure service that delivers high availability and network performance to applications.

0 comments No comments

1 answer

Sort by: Oldest
  1. AI answer

    2026-11-15T20:40:53.23+00:00

    Configure the Azure FortiGate pair as IPsec responders behind the internal and public Standard Load Balancers; use the ILB private frontend for ExpressRoute tunnels and separate ELB public frontends for Internet tunnels, rather than expecting Azure-initiated tunnels to select those frontend IPs automatically.

    1. ExpressRoute tunnel through the internal load balancer

    For a tunnel initiated on-premises:

    1. Configure the FortiGate pair in the ILB backend pool.
    2. Create a static private frontend IP on the ILB.
    3. Configure an HA Ports rule:
      • Protocol: All
      • Frontend/backend port: 0
      • Backend pool: both Azure FortiGates
      • Health probe: a port answered only by the active FortiGate
    4. Configure the on-premises FortiGate’s remote gateway as the ILB frontend private IP.
    5. Make the Azure FortiGates responders for this tunnel.

    HA Ports handles all TCP and UDP ports and is intended for firewalls, VPNs, and SD-WAN appliances. The health probe directs new flows to the healthy FortiGate. ^1^

    If more than one ILB frontend must reach the same backend pool, every corresponding HA Ports rule must have Floating IP enabled. ^2^

    Be aware that Azure guarantees HA Ports flow symmetry only under the documented single-ILB/backend configuration constraints. It does not guarantee symmetry across multiple load balancers or multiple frontend configurations. Validate the combined ILB/ELB topology against Fortinet’s supported HA template before production deployment. ^2^

    2. Two Internet tunnels

    A public Standard Load Balancer can have two frontend public IP configurations pointing to the same FortiGate backend pool. Each frontend needs its own load-balancing rule for IPsec traffic. When different frontends target the same backend port, enable Floating IP on each rule and configure the FortiGate backend to accept the preserved frontend addresses as required by the Fortinet architecture. ^3^

    The intended packet flows are therefore:

    • On-prem tunnel 1 → Public FIP 1 → active Azure FortiGate
    • On-prem tunnel 2 → Public FIP 2 → active Azure FortiGate

    Both frontends should target the HA backend pool, not permanently target Azure FortiGate 1. The probe then moves traffic to FortiGate 2 when it becomes active. This directly matches the FortiGate active/passive tunnel design described in the corresponding case. ^4^

    3. Azure-initiated tunnels and outbound public IPs

    A load-balancing rule is inbound-only; it does not automatically make Azure FortiGate-originated traffic use that rule’s frontend IP. ^1^

    A public Standard Load Balancer outbound rule can SNAT traffic from backend VMs to selected frontend public IPs. Configure this under:

    Azure portal → Load balancer → Outbound rules → Add

    Select:

    • Backend pool containing the FortiGate NICs
    • UDP protocol where appropriate
    • Required frontend IP configuration
    • Manual SNAT-port allocation
    • Disable outbound SNAT on overlapping inbound rules before assigning their frontend IPs to an outbound rule

    Outbound rules apply only if the VM does not have an instance-level public IP. ^5^

    However, attaching FIP 1 and FIP 2 to one outbound rule does not give FortiGate deterministic per-tunnel selection. Azure uses the available frontend SNAT resources; the two IPs primarily form an outbound IP set and increase SNAT capacity. ^5^

    Therefore, for predictable IPsec peer identity:

    • Prefer both on-premises devices initiating their tunnels to FIP 1 and FIP 2.
    • Configure the Azure FortiGate as responder.
    • If Azure must initiate each tunnel from a particular public IP, use a Fortinet-supported SDN Connector/IP ownership failover design rather than assuming two ELB frontend addresses behave like two independent FortiGate WAN interfaces. ^6^

    4. Physical Internet-path diversity

    Two public frontend IPs on one load balancer provide two logical tunnel destinations, but they do not by themselves establish two independently controlled physical ISP paths.

    For Azure-side infrastructure resilience:

    • Use Standard, zone-redundant public IPs and a zone-redundant load balancer.
    • Place the two FortiGate VMs across availability zones where supported.
    • Keep both public IPs and tunnels, but treat them as endpoint/tunnel redundancy rather than proof of physical carrier diversity. Azure recommends zone-redundant frontend public IPs to protect the load-balancer data path from a zone failure. ^7^

    For the ExpressRoute path, genuine physical diversity requires redundant paths through different providers or peering locations; two connections at one peering site do not provide site diversity. ^8^


    References

    1. Azure Load Balancer components
    2. High availability ports overview
    3. Multiple frontends - Azure Load Balancer | Microsoft Learn
    4. SDWAN Tunnel on Azure FortiGate active - standby using load balancer - Microsoft Q&A
    5. Outbound Rules for Azure Load Balancer | Microsoft Learn
    6. HA for FortiGate-VM on Azure | FortiGate Public Cloud 7.4.0 | Fortinet ...
    7. Azure Load Balancer Best Practices
    8. Design and architect Azure ExpressRoute for resiliency
    AI-generated content may be incorrect. Read our transparency notes for more information.

    Was this answer helpful?

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.