Thanks for your reply.
I tried to turn off all the physical machines and start them up, and now they are back to normal.
During this time, there was no human intervention.
May I ask what things I need to pay attention to, such as the log.
It turned out that the cluster was fine, but when it was shut down and restarted, it broke.
When the corosync process was started, the QM command was stuck.
I have confirmed that the time synchronization server is working properly.
I found these articles, but they don't seem to...