A limited number of customers may be experiencing high connection latency and high CPU utilization during connection establishment in Azure Database for PostgreSQL Single Server. The impact is more visible in servers with a high rate of new connections and/or servers that do not use connection poolers.
We have determined that a recent deployment introduced a regression causing high CPU utilization and connection latencies for a limited subset of PostgreSQL customers. To mitigate the issue, we are working on rolling back the recent deployment in all regions deployed.
The rollback deployment was completed in test environments with no issues observed. The rollback deployment has started in production regions and has completed successfully for over 50% of the affected regions. The deployment is being performed in batches of regions during off-business hours to minimize impact, following our safe deployment practices. The rollback is estimated to be completed by the end of day on 15 Oct 2022 in all Azure regions globally.
During the rollback deployment, all Postgres servers in that region may experience a restart leading to brief unavailability of server (60-120 seconds) for your end users. The rollback deployment including server restarts are carefully monitored by the engineering team.
As we continue to complete the rollback in each region, customers will start seeing the recovery earlier than 15th October depending on when our rollback in that region was finished. We confirm that everything is going as planned and we will continue to monitor the progress of rollback and will provide an update on the mitigation workstream within 12 hours or as events warrant.