We screwed up twice, and had one planned outage for a total “downtime” of less than 7 minutes for the year and partial registration impact affecting about 15% of customers for about 115 minutes. This is an improvement from the 14 hours and 36 minutes, or 99.933% covered in our 2012 Availability Report. We feel that Availability is one of the major things our customers are expecting from us. We report on any availability issues immediately and do a bi-annual review for transparency reasons, as well to keep our own feet to the fire.
Here were our 3 issues that impacted availability for 2013 (only one shows in the chart at the top of this blog post since that is the only time the whole site was down in September for 7 minutes):
June 2013 – 5 Transactions – We had a bad deployment related to the queueing software we use. There was no downtime per se, but we did disappoint 5 people who were trying to register during that time. We have made an improvement so this does not happen again.
September 2013 – We rolled out our new Financial 2.0 capabilities in the end of September. We had to shut down the database and start a new instance. This caused the site to be unavailable for about 7 minutes. While our design has let us do over 2,000 deployments this year with only that one interruption of service, we do strive to minimize having any downtime as we continue to rapidly improve our system on a continuous basis. We think this is a strategic advantage for us and our customers.
November 2013 – We botched this one and impacted about 15% of our customers for about 2 hours on a Sunday afternoon. You can read a full report here, but the issue was self-inflicted by the person who was doing weekend support and changed a configuration parameter. We did not realize the impact, but were able to fix it quickly once we realized the issue. We have put safeguards in place to make sure this does not happen again.
Hopefully we are not jinxing ourselves by posting this a week before the year end…