On 2008-10-23T15:36:04, Xinwei Hu <[EMAIL PROTECTED]> wrote:
> Hi Lars,
>
> Thanks for the review. ;) Below is a little explain about your concerns.
Hi Xinwei, thanks, I merged this RA too.
> Yes, the nfsserver have to stack with IPaddr.
> For the newly started client, you either have a error of "Program not
> registered" when nfsserver not started yet, or a error of "No route to
> host" when IPaddr not started yet. I think there's no difference from
> the client's point of view ;)
Right. I am wondering if we need an additional mechanism though -
starting the IP up in "passive" mode (ie, one which won't send rejects
but discards all in-bound traffic silently), and then only activate it
after the service has become operational.
> For the already connected clients, it seems the clients will hang
> until the server notifies the change.
Unless they do happen to send a message in that brief intervaland get a
reject, I think.
> > I wonder where that snippet comes from; do we actually have a debug
> > library anywhere? I've seen it in one or two RAs, but just curious.
> Me don't know either. I use a template to write resource agent, and
> these lines are in the template for a long time already. ;)
Hmmm, maybe the template needs fixing ;-)
> > If this was called "ip" instead of "nfs_ip", it could reference the
> > IPaddr parameter in the CIB (possible with pacemaker-1.0) and only be
> > specified once.
> >
> > (We probably want to eventually have hints like that for the GUI to
> > automatically support this ...)
>
> Interesting. I haven't noticed that change, and thought we must assign
> the refid manually. What if there're 2 IPaddr in the group then ?
You have to specify the refid manually (though the GUI could of course
auto-deduce this), but the parameter name you'd get would still be
"ip".
My thinking was that we should augment the metadata with a "Hey, and I
very likely need to inherit an IP address", and then the (G)UI would say
"Please add an IP address dependency", and default to using refid to
inherit that - not a bug of any sort in the current code, but might be a
worthwhile addition.
Regards,
Lars
--
Teamlead Kernel, SuSE Labs, Research and Development
SUSE LINUX Products GmbH, GF: Markus Rex, HRB 16746 (AG Nürnberg)
"Experience is the name everyone gives to their mistakes." -- Oscar Wilde
_______________________________________________________
Linux-HA-Dev: [email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha-dev
Home Page: http://linux-ha.org/
