On Thu, Jan 08, 2009 at 11:55:42AM +0100, Andrew Beekhof wrote:
> On Wed, Jan 7, 2009 at 16:35, Dejan Muhamedagic <[email protected]> wrote:
> >> > 3. New "prefer" statement/clause for location preference
> >> >
> >> > old:
> >> > location ms-drbd0-master-on-xen-1 ms-drbd0 rule role=master 100:
> >> > #uname eq xen-1
> >> > new:
> >> > prefer ms-drbd0:Master xen-1 # score 100 is implied
> >>
> >> i really wouldn't like to see a score being inferred, but otherwise i
> >> like the new syntax
> >
> > Score of 100 is what is typically used.
>
> But for no good reason.
> Its just the number I pulled out the air when I wrote some of the
> initial examples.
>
> Its no better, nor more likely to be correct, than any other value.
Of course not. It's just a nice round value to which most humans
react in a certain way. My guess is that say 117 or 53, though
there's nothing inherently wrong with them, would provoke a
different kind of emotions, if it's not preposterous to talk
about emotions here.
> > It is also good to have a
> > "standard" value, IMO. Anyway, if one wants to use another value,
> > the location command is of course still available.
>
> But then the user misses out on the rest of the improvements.
>
> Bottom line, the user should _always_ specify a score for these sort
> of constraints/preferences.
As I said before, most people probably just care about having a
"primary" and a "secondary" (or standby) node. For them one kind
of preference is enough. I was trying to address needs of the
majority.
At any rate, after discussing this further on the IRC, we agreed
to have sth like:
prefer rsc[:role] node [score]
where score would default to 100 if not specified.
Dejan
> _______________________________________________
> Linux-HA mailing list
> [email protected]
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
