Log rotate for compute nodes

[poquirion]

A log rotate script should be added for compute nodes on the managment node. It can be problematic for long running clusters:

a rotate file with something like this listing all the possible compute node would do the trick.

/var/log/<compute-prefix1>-*/*log {
[...]
}
/var/log/<compute-prefix2>-*/*log {
[...]
}
[...]