We had about 2 minutes at 7:25 AM Eastern this morning where about half of requests were not being serviced properly. No transactions were lost, but users would have had to click refresh on their browsers.
There was an unusual bug on one of our http servers where requests were not properly rerouted to the backup automatically as it should have been – hence the 2 minute delay. The design is supposed to catch this type of error so users would not see any failed requests.
The issue was the first we have had in 2014. That is 2 minutes of partial downtime our of about 180,000 minutes so far this year for a 99.998% uptime.