I'm solving time/date synchronization of my small cluster, and I'm not
sure which solution is the best. I'm using Debian on 1 master/8 diskless
nodes with root over NFS. Filesystems for nodes are created via
"diskless" package. I just figured out how to synchronize master from
external source via ntp protocol. Now the problem is how to synchronize
diskless nodes with master. Cluster is used fulltime for time-consuming
computations and after several days/weeks/months of uptime the dates of nodes
could quite differ.
I didn't have any problems with unsychronized dates but I'm not sure if
such things like files created in future etc. couldn't be source of some
nasty bugs. I'm not sure whether running ntp server on master and ntp
synchronization on nodes is appropriate solution (maybe it is too heavy
weight), other possibility is installing rdate and adding simple
start-up and cron scripts. However I don't want to reinvent the wheel
and I would like to hear how other people solve this problem.