[Pacemaker] A function demand of the new environment.

Lars Ellenberg lars.ellenberg at linbit.com
Mon Sep 21 08:04:48 EDT 2009


On Mon, Sep 21, 2009 at 12:00:37PM +0200, Andrew Beekhof wrote:
> On Fri, Sep 18, 2009 at 3:09 AM,  <renayama19661014 at ybb.ne.jp> wrote:
> > Hi Andrew,
> >
> >> I still don't understand why you can't simply make them regular
> >> cluster resources.
> >> Can you please explain?
> >
> > I understand your opinion.
> > I surely understand that simple RA can operate an object of respawn by shifting.
> >
> > However, an object of respawn of all users cannot shift to RA.
> > For example, there is the user who does respawn of processing to work even if a node becomes the
> > standby state.
> 
> I'm almost afraid to ask, why would someone need this capability?
> 
> > But, a primitive resource and the Clone resource stop when a node becomes the standby state.
> > The user needing this respawn cannot easily shift to new environment.
> 
> True.

If someone wanted to respawn some process, regardless of
online/standby/whatever state of the cluster manager,
that process is independend from the cluster manager.

Then why use the cluster manager configuration file at all?

On Linux (and other unix-ish systems), put it into inittab.

If you prefer indirections (for administrative or other purposes),
respawn some wrapper script from your inittab (just in case the wrapper
script dies unexpectedly), and put other respawn lines into some
config file that is parsed from that wrapper script.

Problem solved.

-- 
: Lars Ellenberg
: LINBIT | Your Way to High Availability
: DRBD/HA support and consulting http://www.linbit.com

DRBD® and LINBIT® are registered trademarks of LINBIT, Austria.




More information about the Pacemaker mailing list