[Linux-HA] instance_attribute IDs, crm_resource, GUI

[Andrew Beekhof]

Right now, in cvs, crm_resource will do its very best to figure out
what to update.

If you just give a name and there is no existing entry, it will create
one with set name = resource name, and ID = setname+"-"+attribute.
If you just give a name and there is exactly ONE existing entry, it
will update that.
If you just give a name and there are multiple existing entries, it
will show you them and tell you to pick one.

If you give a name and a set-name and there is no existing entry, it
will create one with ID = setname+"-"+attribute.
If you give a name and a set-name and there is exactly ONE existing
entry, it will update that.
If you give a name and a set-name and there are multiple existing
entries, it will show you them and tell you to pick one.

If you give a name and an ID and there is no existing entry, it will
create one with set name = resource name.
If you give a name and an ID and there IS an existing entry, it will
update that.
By definition there can be at most one entry with a given ID.

If you give a name, a set-name and an ID and there is no existing
entry, it will create one.
If you give a name, a set-name and an ID and there is an existing
entry, it will update that.
By definition there can be at most one entry with a given ID.

I hope this will be the last we need to cover this :-)

On 7/27/06, Lars Marowsky-Bree <lmb at suse.de> wrote:
> On 2006-07-26T18:29:16, Andrew Beekhof <beekhof at gmail.com> wrote:
>
> > In which case?
> > How can we know if the admin wants to set a new value or change an existing
> > one?
> > What if the only set that exists only applies on weekends and today is
> > Tuesday?
> >
> > Basically I disagree that we can make a good guess.
>
> If that is true, we mustn't guess at all and refuse to do anything w/o
> full specification of the attribute set, which is also cumbersome.
>
> I'm quite sure there's at least two cases where we _can_ make a guess:
>
> - No instance/meta_attributes at all: Create one.
> - Only one, but w/o any rules: Modify that one.
>
> That, I think, should cover 95% of the cases. I actually don't care if
> we have to run a simulation or not, just that we don't surprise our
> users ;-)
>
> And, then:
>
> - Anything else (ie, several attribute sets, constrained with rules,
>   whatever): Refuse to operate w/o specification of the id.
>
>
> Sincerely,
>     Lars
>
> --
> High Availability & Clustering
> SUSE Labs, Research and Development
> SUSE LINUX Products GmbH - A Novell Business     -- Charles Darwin
> "Ignorance more frequently begets confidence than does knowledge"
>
> _______________________________________________
> Linux-HA mailing list
> Linux-HA at lists.linux-ha.org
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>