Our primary database was briefly taken offline (less than 1 minute) while deploying a change related to collecting performance metrics. The change was quickly rolled back and normal service resumed.