The team worked diligently to implement a fix and made several improvements to restore full access. By 17:30 PST, the service was back up and running as expected, with close monitoring to ensure full service. The team also planned to publish a public postmortem to explain the incident and outline measures to prevent similar issues in the future.
Key takeaways:
- An elevated level of API errors was experienced and was under investigation.
- The issue was identified to be a problem with the database replicas, impacting ChatGPT and non-completion API endpoints partially, and completion API endpoints minimally.
- A fix was being implemented and several improvements were in progress to restore full access.
- Service was restored and is being closely monitored to ensure full service and prevent similar issues in the future.