[torqueusers] torque daemons can not be stopeed duringuninstall

Justin Finnerty justin.finnerty at uni-oldenburg.de
Wed Dec 3 07:27:58 MST 2008


Hi all 
 
> My vote goes to leaving the shutdown procedures of daemons to
> administrators, as Garrick suggests.  In this way, sites know that
> there was a conscious decision to remove torque and not simply an
> accidentthat occurred.  Is it so difficult to shutdown the daemons
> prior to using RPM to remove the package?  I don't know how your
> configuration is setup, but in ours, it is as simple to do a two step
> process as a one step process to remove TORQUE.

I manage a cluster using the bcfg2 configuration management tool.
Having used the system for nearly two years now, I do not know how
anyone can effectively manage a cluster without a similar tool!  This
means that the management of software packages is no longer performed
directly by the administrator.

With bcfg2, to update software you drop the new RPM into the
configuration server. The cluster nodes then update themselves against
the configuration server. This can occur at any time, and for nodes that
are away for repair it can be at a much later date.

Also, as we may need to update tens to thousands of nodes in a cluster
which really means we can not rely on human intervention.  Therefore
when we make RPMs of services intended for deployment on a cluster we
have to be more careful with making the RPM.

My $.02

Cheers
	Justin


-- 
Dr Justin Finnerty
Rm W3-1-165         Ph 49 (441) 798 3726
Carl von Ossietzky Universität Oldenburg



More information about the torqueusers mailing list