Machines are back in the cluster after investigation.
The issue was related to corrupted files storing configurations.
Seven configurations have been impacted.
Posted Dec 08, 2017 - 11:08 CET
Monitoring
We isolated two machines that were sending 521 errors during this period.