> > I've been struggling for a couple of weeks to figure out
> > how to get my node to failover when I unplug the network
> > cable. I've read everything I can find
> > and I'm still struggling.
> >
> > I'm running
> > Heartbeat-2 2.1.2
> > Heartbeat-Gui(hb_gui) 2.1.2
> > Kubuntu 7.10 with AMD64 kernel
> >
> > I'm trying to set up a simple Active-Passive cluster.
> >
> > Most things seem to behave well, except for pingd. That
> > is, if I manually kill services from the bash shell (e.g.
> > ifdown eth0 ), my resources will
> > failover nicely. But when I yank the network cable,
> > nothing happens.
> for pingd to work, you need to have it running on all the nodes
> (either as a clone or using a respawn directive)
Arghh! This is driving me insane. I simply cannot get pingd to
play nicely with the gui. If anyone has a success story, I
would be exceedingly grateful to hear it.
I have tried both clones and respawn, and the I cannot get my
resources to failover reliably or predictably. The GUI
absolutely does not like "defined" or "not defined" operators.
These are supposed to be specified without values in the cib.xml
file. But the GUI will not accept any operators without values.
Now I am faced with deciding on a fall back position
(1) Go to semi-automatic mode. If one of my nodes goes dead,
someone will have to figure it out and then
manually make them failover. This will probably remove at
least a 9 or two from my uptime.
(2) Abandond the gui altogether and become friends with the
command line tools. I have played around with these a bit
and it is really difficult to get meaningful information
and figure out how my cluster is behaving.
I am currently leaning towards 1. Since then If I am not around
when things fail, I can talk someone through what to do over the
phone with the gui.
> > Because I am one of the few Linux users in my organization,
> > I have a strong
> > desire to be able to administer everything from hb_gui. The
> > Quickstart examples for pingd in http://www.linux-ha.org/pingd
> > cannot be used with hb_gui, because (1) it does not allow
> > boolean_op="or" (2) it does not allow
> > score_attribute or to add an attribute (defined) without a value.
> many things are not possible with the gui :(
> > So I tried to add an rsc_location constraint with a high score (300)
> > attribute="pingd" operation="defined" value="true". I also added a
> > monitor
> > for pingd. I'm a little unclear on whether I need the respawn
> > directive in
> > ha.cf or can it be handled by the resource? I've tried it both ways.
>
> either will work
>
> > I wish there was a way to figure out what pingd was doing. I have
> > verified,
> > via bash shell, that I cannot ping my ping node when I disconnect
> > the cable.
> > But crm_failcount
>
> failcount has nothing to do with it
>
> > and cibadmin -Ql don't seem to indicate that anything is
> > wrong.
>
> from the configuration you attached, it doesn't look like
> either node
> can talk to the ping node.
>
>
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems
