The gray question mark indicates that the daemon responsible for statistical and health updates was unable to talk to that node's software. It can also indicate that a subsystem on that node is not responding. The situation is often caused by unstable storage connectivity.
To determine the cause in your particular case you need to review the logs on both the "primary" node (where you are logged in) and on the "failed" node.
"journalctl -b" is a good start