Hi,

On Thu, Jan 08, 2009 at 06:07:08AM -0800, Joe Bill wrote:
> 
>  On Thu, Nov 20, 2008 at 08:32:52AM -0700, Ryan Kish wrote: 
> >> - Both nodes have a two ldap configurations, one for running as the
> >> master, one for running as the slave. I had to write a couple of OCF
> >> scripts for each service (master and slave). 
> 
>  On 2008-11-20, Dejan Muhamedagic <dejanmm[at]fastmail.fm> wrote:
> >
> > I guess that this could be better handled by a multi-state
> > (master-slave) resource. For which you'd need an appropriate OCF
> > RA. Basically, what you already have could be combined in a
> > single RA, with two extra operations (actions): promote (slave
> > to master) and demote (vice versa). Your configuration would
> > also be simpler. 
> > 
> > Dejan
> 
> I've recently given multi-state RAs a hard, heavy thought.

Probably they needed it ;-)

> I have to disagree on several issues brought up above:
> 
> - the "master/slave, promote/demote" is not a universal
> multi-state model and appears, to me at least, to poorly
> represent instance relationships between several different
> components/applications.

What is exactly missing? Can you please be more specific. And
provide perhaps some examples.

> - the mysql replication model as a master-slave architecture,
> by giving the "master"/"slave" moniker values to an option
> determining a hosts's role, is a misrepresentation. The fact of
> the matter is that one host can run both, the master AND the
> backup instances at the same time, and all still make sense and
> be coherent with the HA supervisor, depending on the
> availability/reachability of nodes. I have yet to figure the
> advantage of           changing on the fly the state of an
> instance, particularly in it's relationship with other
> instances, whether the same or different
> components/applications are involved, instead of just stopping
> and restarting an instance with a different configuration.
> 
> It makes a lot more sense to use the supervisor's logical
> instance control facilities than hardcode in the RA

Can you define: "supervisor", "logical instance", and "control
facilities".

> promote/demote rules that have to track and manage the states
> of multiple instances, and especially difficult to write
> "decision" code that requires to assess and control the states
> of several different compnents.

Several? I thought one RA would control only one resource.

Thanks,

Dejan
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to