AWS Global Accelerator handles failovers for dual-stack endpoints by continuously monitoring the health of all endpoints within an endpoint group and routing traffic only to those that are healthy. Dual-stack endpoints support both IPv4 and IPv6 and can be Network Load Balancers, Application Load Balancers, or Amazon EC2 instances that have a primary IPv6 address attached[3][9][6].
Here is how failover works in detail for dual-stack endpoints:
- Global Accelerator performs health checks on all endpoints in an endpoint group. If an endpoint becomes unhealthy, it stops routing traffic to that endpoint and instead routes traffic to other healthy endpoints within the same group[3][5].
- If no healthy endpoints with a weight greater than zero exist in the current endpoint group, Global Accelerator attempts to fail over to a healthy endpoint in another endpoint group (typically in a different AWS Region). This failover ignores the traffic dial settings, meaning even endpoint groups with traffic dial set to zero are considered during failover attempts[5].
- If no healthy endpoints with a weight greater than zero are found after checking the three closest endpoint groups, Global Accelerator "fails open" by routing traffic to a random endpoint in the closest endpoint group to the client, regardless of health status[5].
- When the previously unhealthy endpoints recover, Global Accelerator resumes normal routing behavior, directing traffic back to healthy endpoints with traffic dials above zero. However, existing active connections continue to route to their original endpoints until reset[5].
- Dual-stack accelerators require that all endpoints be dual-stack as well. When updating an accelerator to dual-stack, Global Accelerator ensures all endpoints have primary IPv6 addresses and adds them if necessary to EC2 instances[6][9].
- Global Accelerator provides static dual-stack IP addresses as a single point of contact, routing client traffic over the AWS global network to the closest healthy dual-stack endpoint, thereby improving availability and performance[3][10].
- Client IP address preservation can be enabled for some dual-stack endpoints, but there are restrictions, such as disabling this feature for dual-stack Network Load Balancers in IPv4 accelerators[3].
In summary, AWS Global Accelerator manages failover for dual-stack endpoints by health monitoring, weighted routing, regional failover, and fallback to nearest endpoints, all while requiring consistent dual-stack support across endpoints to ensure seamless IPv4 and IPv6 traffic handling and high availability[3][5][6][9].
Citations:
[1] https://docs.aws.amazon.com/global-accelerator/latest/dg/introduction-how-it-works.html
[2] https://docs.aws.amazon.com/global-accelerator/latest/dg/about-endpoints.html
[3] https://docs.aws.amazon.com/global-accelerator/latest/dg/about-endpoints-caveats.html
[4] https://docs.aws.amazon.com/global-accelerator/latest/dg/disaster-recovery-resiliency.html
[5] https://docs.aws.amazon.com/global-accelerator/latest/dg/about-endpoints-endpoint-weights.unhealthy-endpoints.html
[6] https://aws.amazon.com/blogs/networking-and-content-delivery/global-accelerator-ipv6-ec2/
[7] https://docs.aws.amazon.com/pdfs/global-accelerator/latest/dg/global-accelerator-guide.pdf
[8] https://aws.amazon.com/global-accelerator/faqs/
[9] https://docs.aws.amazon.com/global-accelerator/latest/dg/about-endpoints-adding-endpoints.html
[10] https://aws.amazon.com/global-accelerator/features/