good work

good for commit
On Tue, 2008-08-12 at 08:46 +0200, Fabio M. Di Nitto wrote:
> Hi Steven,
> 
> new patch in attachment:
> 
> - change list_add to list_add_tail
> - fix unloading of services to respect original behaviour
>    (last loaded, first unloaded)
> - few whitespace cleanups
> 
> Fabio
> 
> On Mon, 11 Aug 2008, Steven Dake wrote:
> 
> > I don't have a problem with this patch but it breaks shutdown.
> >
> > Example:
> >
> > Before patch (pressing ctrl-c)
> > Aug 11  8:21:18.601215 [SERV ] Unloading openais component:
> > corosync_confdb v0
> > Aug 11  8:21:18.601259 [SERV ] Unloading openais component: corosync_cpg
> > v0
> > Aug 11  8:21:18.601287 [SERV ] Unloading openais component: corosync_cfg
> > v0
> > Aug 11  8:21:18.601314 [SERV ] Unloading openais component: corosync_evs
> > v0
> > A
> > After patch (pressing ctrl-c)
> > Aug 11  8:19:20.983285 [TOTEM] entering OPERATIONAL state.
> > Aug 11  8:19:23.045472 [SERV ] Unloading all openais components
> > Aug 11  8:19:23.045550 [SERV ] Unloading openais component: corosync_evs
> > v0
> > Aug 11  8:19:23.050464 [SERV ] Unloading openais component: corosync_cfg
> > v0
> > Aug 11  8:19:23.050516 [SERV ] Unloading openais component: corosync_cpg
> > v0
> > Aug 11  8:19:23.050544 [SERV ] Unloading openais component:
> > corosync_confdb v0
> > Aug 11  8:19:23.050754 [MAIN ] AIS Executive exiting with status -1 at
> > main.c:151.
> >
> > The original behavior must be retained in the unloading of all the
> > modules, that is, The last loaded module should be unloaded first or it
> > will break other services possibly pacemaker.
> >
> > regards
> > -steve
> >
> > On Mon, 2008-08-11 at 17:01 +0200, Fabio M. Di Nitto wrote:
> >> Hi guys,
> >>
> >> there is a substantial problem in the way keys and objectes are created
> >> within the objdb.
> >>
> >> Let's assume this config file:
> >>
> >> <foo>
> >>   <bar name="1" action="off"/>
> >>   <bar name="2" action="on"/>
> >> </foo>
> >>
> >> (very similar to our fence method configuration snippet)
> >>
> >> Using list_add to create the keys and objects, the configuration in the
> >> object db is mirrored:
> >>
> >> <foo>
> >>   <bar action="on" name="2"/>
> >>   <bar action="off" name="1"/>
> >> </foo>
> >>
> >> the patch in attachment change list_add into list_add_tail to insert data
> >> in the objdb in the same order as they have been pushed.
> >>
> >> While _generally_ this shouldn't be an issue, we have services like fence,
> >> that relies on the correct order of data to be returned from the objdb in
> >> order to perform their operations correctly.
> >>
> >> Please apply
> >>
> >> Fabio
> >>
> >> --
> >> I'm going to make him an offer he can't refuse.
> >> _______________________________________________ Openais mailing list 
> >> [email protected] 
> >> https://lists.linux-foundation.org/mailman/listinfo/openais
> >
> 
> 
> --
> I'm going to make him an offer he can't refuse.

_______________________________________________
Openais mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/openais

Reply via email to