On Mon, Feb 15, 2010 at 12:37 PM, Steven Dake <[email protected]> wrote:

> On Mon, 2010-02-15 at 12:35 -0700, hj lee wrote:
> >
> >
> > On Fri, Feb 12, 2010 at 10:57 PM, Steven Dake <[email protected]>
> > wrote:
> >         On Fri, 2010-02-12 at 15:25 -0700, hj lee wrote:
> >         > Hi,
> >         >
> >         > I set the nodeid in corosync.conf. But this nodeid is
> >         ignored when the
> >         > interface is bound to 127.0.0.1, and nodeid becomes
> >         16777343. This new
> >         > nodeid confuses the pacemaker. The pacemaker thinks this is
> >         new node.
> >         > So the corosync is connected, but the pacemaker thinks the
> >         peer is
> >         > offline. To fix this issue, the localhost.nodeid should be
> >         initialized
> >         > to the configured nodeid in totemudp_initialize().
> >         >
> >         > Patch:
> >         > +localhost.nodeid = instance->totem_config->node_id;
> >         >
> >
> >
> >         Ya that is a good suggestion if node_id is defined in the
> >         configuration
> >         file.  Could you work up a patch?
> >
> >         regards
> >         -steve
> >
> > I attached the patch against flatiron. It's just one line addition.
> >
> > Thanks
> > hj
> >
> >
> HJ,
>
> thanks for the contribution.  I have a question about behavior though.
> If nodeid is not specified in the totem configuration,
> totem_config->node_id is set to 0.  node_id's of 0 are not allowed in
> totem.  If the user doesn't set node_id, and has your patch applied,
> wouldn't totem assert in various places?  I believe you want to test
> that the value of node_id is nonzero before assigning.
>
>
This is handled by totemip_iface_check(). If nodeid is zero, then
totemip_iface_check() will use the IP address of binding net. So it's OK to
set zero to nodeid. Also this is exactly same as how
totemudp_instance->totem_interface->bindnet.nodeid is set.

Also the patch uses spaces.  We require tabs for indentation.  Have a
> look here at our coding style guidelines:
>
> http://www.corosync.org/doku.php?id=dev:contributor_guide
>
> I updated the patch using tab.

Thanks
hj


-- 
Peakpoint Service

Cluster Setup, Troubleshooting & Development
[email protected]
(303) 997-2823
Index: exec/totemudp.c
===================================================================
--- exec/totemudp.c	(revision 2651)
+++ exec/totemudp.c	(working copy)
@@ -1777,6 +1777,7 @@
 	instance->totemudp_target_set_completed = target_set_completed;
 
 	totemip_localhost (instance->mcast_address.family, &localhost);
+	localhost.nodeid = instance->totem_config->node_id;
 
 	/*
 	 * RRP layer isn't ready to receive message because it hasn't
_______________________________________________
Openais mailing list
[email protected]
https://lists.linux-foundation.org/mailman/listinfo/openais

Reply via email to