Hello, the issue has been identified.
The cause lies in the server operating system's automatic security updates: Ubuntu automatically installs security patches in the early morning every few days, and the installation process restarts core cluster services. During the restart, the internal DNS component of the cluster is reverted to a default image that cannot be pulled from the domestic network, causing the DNS component to fail to start. Once internal domain name resolution fails, Halo cannot connect to the database, resulting in a 502 error. This is also why the console does not always indicate that the Zeabur service is abnormal—the platform's own components are actually fine; the issue lies with the cluster's internal DNS.
"Reinstalling the Zeabur service" reverts the DNS component back to a mirror source available domestically, which is why it recovers after every reinstallation, but it recurs during the next automatic system update.
We have made two adjustments to your server: first, we configured the system to stop automatic updates from restarting cluster services, and second, we pinned the DNS component configuration to prevent it from being reverted. You should no longer experience this periodic 502 error, and there is no need to manually reinstall it anymore. If you encounter any further issues, please feel free to reply at any time.
Forum