Today we had an incident that lasted for approximately one hour in the NYC region. I wanted to take the time to address this incident and review what happened, why it happened, what we did to resolve it, and what we’re doing to prevent future issues like this.
The incident began at 11:00 a.m. ET when some customers in the New York City region experienced high latency over our DNS1 Anycast network.
The cause of this incident stemmed from an issue with one of our hosting providers in which a circuit was inadvertently turned off. This resulted in DNS traffic to the Washington, D.C. region to be rerouted to servers in New York City. While this traffic shift is what we expect BGP to do in instances like this, the load was too great to resolve some requests in a timely manner. This resulted in the latency issue experienced by some of our customers.
Prior to receiving our first ticket, our team had already recognized the issue was with that particular upstream provider. We had visibility into the traffic moving from DC to NYC, and we were able to fully diagnose and solve the issues by 12:10 p.m. ET—24 minutes after our first support ticket was received.
But we’re committed to doing better. We’ve already had plans to increase our load capacity, including the NYC region, which will allow us to have greater resilience in the case of unbalanced loads resulting from unplanned traffic shifts. Our goal is to serve our customers and provide the best possible experience.
A while ago, I promised transparency, action, and accountability. I’m here to keep that promise.