Odisconnect
Background of the incident: Within Convex, we have set up two projects, named Axon and Axon-dev. Axon is used for the official runtime environment, while Axon-dev is for internal testing purposes. The issue arose when the code for Axon and Axon-dev was completely identical, yet two different phenomena occurred.
Phenomena observed: At the time, there were over 10 users using the system, including 2 of our testers. These testers simply logged into the website and did not perform any actions. However, the page would suddenly crash. Upon opening the browser console, an error was observed: the connection to the Convex service was interrupted. Meanwhile, the testers also accessed Axon-dev without encountering the same issue.
Subsequently, developers started to check the logs in Convex for troubleshooting and discovered the following errors in the Axon project, while the Axon-dev project did not exhibit any
Regarding the official response and proposed solution: We were advised to reduce the use of complex algorithms in our queries and mutations. We will try our best to address this. However, we still want to understand the current issue's cause. Why is it that the testers, without any operations, would get disconnected from Convex, while some users can maintain normal communication despite frequent operations?
Phenomena observed: At the time, there were over 10 users using the system, including 2 of our testers. These testers simply logged into the website and did not perform any actions. However, the page would suddenly crash. Upon opening the browser console, an error was observed: the connection to the Convex service was interrupted. Meanwhile, the testers also accessed Axon-dev without encountering the same issue.
Subsequently, developers started to check the logs in Convex for troubleshooting and discovered the following errors in the Axon project, while the Axon-dev project did not exhibit any
Regarding the official response and proposed solution: We were advised to reduce the use of complex algorithms in our queries and mutations. We will try our best to address this. However, we still want to understand the current issue's cause. Why is it that the testers, without any operations, would get disconnected from Convex, while some users can maintain normal communication despite frequent operations?
