[Pacemaker] About "record-pending"

[Yan Gao]

On Thu, 2008-10-30 at 08:09 +0100, Andrew Beekhof wrote:
> On Oct 30, 2008, at 7:56 AM, Andrew Beekhof wrote:
> 
> >
> > On Oct 30, 2008, at 7:13 AM, Yan Gao wrote:
> >
> >> On Mon, 2008-10-27 at 18:22 +0800, Yan Gao wrote:
> >>> On Mon, 2008-10-27 at 10:08 +0100, Andrew Beekhof wrote:
> >>>>>
> >>>>> Another issue is it doesn't trigger cib notify, hence mgmtd  
> >>>>> cannot be
> >>>>> aware of the event.
> >>>>
> >>>> Is the crm sending the update with the cib_inhibit_notify flag?
> >>>>
> >>> If it's LRM_OP_PENDING, it will always be set in:
> >>> te_actions.c:290
> >>> 		call_options |= cib_inhibit_notify; /* We don't want to
> >>> know about these updates */
> >> Andrew, clients could never get the notify because cib_inhibit_notify
> >> is always set here.
> >
> > I'll fix that today.
> 
> Ok, both parts fixed:
>     http://hg.clusterlabs.org/pacemaker/stable-1.0/rev/e823b091e55c
> 
Great, thanks!
> >
> >
> >> And the cib_scope_local is also set, so probably a
> >> pending operation would not be recorded in another node ?
> >
> > No.  That flag just means perform the update locally - it doesn't  
> > mean the change wont be distributed.
> >
OK, I see

> >>
> >>
> >>>
> >>> And should it be like the following at line 289 ?
> >>> -if(LRM_OP_PENDING) {
> >>> +if(status == LRM_OP_PENDING) {
> >
> > that would make sense
> >

Regards,
-- 
Yan Gao
China R&D Software Engineer
ygao at novell.com

Novell, Inc.
SUSE® Linux Enterprise 10
Your Linux is ready
http://www.novell.com/linux