When a command needs to read the configuration file when the cluster is replacing the configuration file, an error message is displayed indicating that the configuration file does not exist. How can I avoid this situation?