On 15 September 2023, between 00:43 UTC and 10:05 UTC, approximately, many customers experienced errors or extreme delays accessing the Web Application, APIv2, and other APIs in our West Europe region. Timeout errors and high latency between client applications and the API would occur during a certain amount of time, and several times in the incident timespan.
The incident response team quickly noticed that we had multiple errors related mainly to SQL database and service connectivity. During the incident, the Azure Management Portal was not working properly, giving errors when accessing several resources in the West Europe region. Multiple alerts were lost due to this Azure issue.
The team engaged Azure Support, and it was confirmed that Azure was investigating a major incident in the West Europe region regarding the networking infrastructure (tracking ID: YKS4-9PZ). This incident started on the 14th September, 23:05 UTC, and impacted several services, including connectivity between services like SQL Databases and App services.
As the majority of data on our platform is stored on multiple SQL Databases and we use Azure to rely on the communication between services, this Azure issue had a severe impact on our services.
Azure team started rolling out a mitigation just after the problem was acknowledged but service connectivity remained unstable until around 9:00 UTC, when full mitigation occurred.
Skills Workflow is highly dependent on several data and networking services from Azure, so any issue with Azure and its service connectivity has a high impact in our service as well.
On this occasion, this Azure infrastructure issue had a big impact on our system availability. Azure tracked this incident (tracking ID: YKS4-9PZ) and has already provided a full RCA on their side (https://app.azure.com/h/YKS4-9PZ/f8e20a).
Our customers rely on Skills Workflow and we understand the impact these incidents have and that's why we are committed to continuously improve our platform.
As next steps, we will try to find ways to reduce the impact these kinds of issues have on our system as well as continue to rely on Azure to continually improve their services' reliability. We believe an effort to improve communication with our clients regarding the impact of failures of this nature and the consequent recovery process is needed. A review of the process on this matter is already in progress. Also in progress is a review and improvement of the alerts process. This review aims to improve incident response time in a more proactive way, also providing better communication with our customers.