From: Karl Vogel <karl.vogel@se...> - 2004-09-20 07:17:59
> I am also interested. What is the best way to run cron ?. I guess it
> is more generic. For me cluster should behave as a single machine
> that means i want my job to be executed only once. So this is
> important with respect to user specific crontab file also . If we run
> cron on each node i guess we will have each cron instance executing
> the same cron job. On the other hand how do we handle job that may
> need to be run on each node. Like taking backup of node specific
> files. Do we need modify all those cron job to be cluster aware and
> run only one instance of cron job inthe cluster. At first look this
> looks simple. But what is the best way ?
For reference.. on Tru64 the Cluster Application Availability (caa)
service merges the cluster cron entries with the 'local node' cron
entries at failover time.
However it's done in a horrible way.. it appends the cluster crontab
file with the local one and adds '#clustercronData' to each line
from the cluster crontab. Resulting in a Tru (pun intended) mess.