I've been working on a robust canary deployment setup for our B2B SaaS platform, and after testing a few approaches, I landed on a combination of Azure Traffic Manager and AWS Application Load Balancer. The goal was to shift a small percentage of our live traffic to a new application version for performance and error rate monitoring before a full rollout.
The setup uses Traffic Manager for geographic routing and failover at the DNS level, pointing to two ALBs—one for the stable version and one for the canary. The ALB then handles the actual traffic splitting based on listener rules, sending, say, 5% of requests to the new target group. I'm curious—has anyone else implemented a similar multi-cloud canary strategy?
I'm particularly interested in hearing about the operational overhead. We're tracking latency differences and error budgets, but I wonder if others have compared the cost-per-unit of this pattern against a single-provider solution using something like App Gateway or a service mesh. The flexibility is great, but does the complexity pay off for your team?
— Jane
— Jane