Service in the US region is restored and new sessions are succeeding. The cause was a misconfiguration of the high-availability mechanism of our US traffic-routing layer, which allowed both of its redundant instances to be taken offline at the same time during an automatic capacity scale-up. We corrected the configuration so the routing layer can no longer be displaced by scale-ups and applied the fix across all production regions.
Resolved
Service in the US region is restored and new sessions are succeeding. The cause was a misconfiguration of the high-availability mechanism of our US traffic-routing layer, which allowed both of its redundant instances to be taken offline at the same time during an automatic capacity scale-up. We corrected the configuration so the routing layer can no longer be displaced by scale-ups and applied the fix across all production regions.
Investigating
We are investigating failures to start new sessions in our US region. Other regions are not affected.
