An Azure service that delivers high availability and network performance to applications.
Gateway Load Balancer (GWLB) is designed for transparent service insertion for inbound/outbound (north–south) flows to workloads that are fronted by a public or internal Standard Load Balancer or public IP. It preserves the original source and destination IPs for those flows, but it cannot be used to insert NVAs into pure east–west traffic between Azure virtual machines.
For east–west inspection, the supported pattern is to place NVAs (or Azure Firewall / Cloud NGFW) in a hub VNet or Virtual WAN hub and steer traffic through them using routing (UDRs or Virtual WAN routing intent), not GWLB:
- Hub-and-spoke with internal Load Balancer + UDRs
- Deploy NVAs in a dedicated subnet in the hub.
- Place an internal Load Balancer in front of the NVAs.
- Use user-defined routes (UDRs) on spoke subnets so that traffic between spokes or between on-premises and Azure is sent to the NVA ILB for inspection.
- This pattern supports east–west inspection and keeps traffic symmetric; the internal Load Balancer ensures both directions of a flow hit the same NVA instance.
- Virtual WAN with Azure Firewall or Cloud NGFW
- In Virtual WAN, deploy Azure Firewall or a third-party NVA such as Palo Alto Networks Cloud NGFW in the virtual hub.
- Use routing intent or Virtual WAN routing policies so that VNet-to-VNet and VNet-to-on-premises traffic is routed via the firewall/NVA for east–west inspection.
- This provides a centralized inspection plane for private traffic (east–west) and can also handle north–south inspection.
- Why GWLB cannot be used for east–west
- GWLB service chaining is supported for inbound traffic to a public load balancer, its return traffic, and outbound traffic from Azure.
- The platform explicitly does not support using GWLB to inject NVAs into east–west traffic between Azure virtual machines.
- For east–west, NVA insertion must be done via routing (UDRs/Virtual WAN) and, where needed, internal Load Balancers in front of the NVAs.
For regulated workloads that require a single inspection plane for both north–south and east–west traffic, the recommended approach is:
- Use a hub-and-spoke or Virtual WAN architecture.
- Centralize NVAs (including Palo Alto or other NGFWs) or Azure Firewall in the hub.
- Use UDRs / Virtual WAN routing intent plus internal Load Balancers (for NVA HA) to steer both north–south and east–west flows through the same inspection plane, while preserving original source IPs at the NVA.
GWLB can complement this for north–south scenarios where transparent service insertion is needed for internet-facing workloads, but it cannot replace routing-based designs for east–west inspection.
References: