[Linux-HA] Heartbeat Shutdown issues

[Andrew Beekhof]

On 10/11/07, Raoul Bhatia [IPAX] <r.bhatia at ipax.at> wrote:
> Andrew Beekhof wrote:
> > On 10/10/07, Raoul Bhatia [IPAX] <r.bhatia at ipax.at> wrote:

> >> as far as i can see, when i issue some "kills" i get a core dump in
> >> /var/lib/heartbeat/cores/hacluster/ - please find it attached.
> >
> > these are only useful on the machine that generated them
> > what i need instead is the stack-trace
> >
> > and why (and to whom) are you issuing "kills"?
>
> good question. i thought i narrowed it down to "heartbeat: master
> control process", as i think it said a couple of times that it waited
> for this pid.
>
> most of the time, after waiting for 5-10 minutes, i simply do a
> "killall heartbeat" to be able to restart heartbeat after some kind of
> test/failure/bug and then manually clean up other processes like lrmd,
> ha_logd and once crmd.
>
> i don't know any other way to be able to restart heartbeat.

_please_ report stuff like this (with logs and "ps axf" outputs^).
its the only way things will ever get better.

^ and if you're inclined, attach to the process and tell us where it
was at the time

btw. do you have "use_logd yes" in ha.cf?  i've seen a number of
shutdowns hang because of it
(i've reported bugs but so far no response)