Hi! Yes, ntp works properly (I know about similary problems if ntp dont working, but looks like this error not about ntp). We tried to restart agents and cm server (include host restart) and even try to install different version (6.1.1) of CM and agents on different hosts, but no luck. And yes, we did "df -h" on the hosts: this shows filesystems used by host. There no errors or some "strange things" in output of the command. We dont use NFS or any other network mounted FS. And I tried as example change directory to nodev fs /run/cloudera-scm-agent/process and do "ls -la" which returns ''." and ".." whitout any delays as expected. Thus, I think, this FS used by cm-agent works properly.
At the end, if we install agent on the same host as Cloudera Manager, then it is ok, no errors about "nodev" filesystems or any other.
Is it a new server or a server that you deleted and recreated with the same name?
try to clean the cache on the cloudera manager server service nscd restart and nscd -i hosts
Are you able to ssh to the server from the cloudera manager server?