[Openais] Problem stoping Corosync+Pacemaker

[Andrew Beekhof]

On Thu, Nov 5, 2009 at 12:59 PM, Adrian Chapela
<achapela.rexistros at gmail.com> wrote:
> Andrew Beekhof escribió:
>>
>> On Thu, Nov 5, 2009 at 11:13 AM, Adrian Chapela
>> <achapela.rexistros at gmail.com> wrote:
>>
>>>
>>> ....
>>>
>>> Madkiss packages init script is the same as I am using.
>>>
>>>>
>>>> This is the stop() function from init/redhat (in the source tarball)
>>>>
>>>> stop() {
>>>>    echo -n $"Stopping Corosync Cluster Engine ($prog): "
>>>>    # If no signal is specified, -TERM is used but _also_ -KILL 3s later
>>>>    # This is far too aggressive for a cluster resource manager
>>>> running on top of Corosync
>>>>    killproc $prog -TERM
>>>>    echo
>>>>
>>>>    echo -n $"Waiting for services to unload:"
>>>>    while
>>>>        pidofproc $prog > /dev/null 2>&1
>>>>    do
>>>>        sleep 2
>>>>    done
>>>>
>>>>    success $"$base shutdown"
>>>>    echo
>>>>
>>>>    rm -f "$lockfile"
>>>>    return 0
>>>> }
>>>>
>>>>
>>>>
>>>
>>> With the above function I have a problem,killproc, pidofproc and success
>>> command doesn't exist on my system (I have changed pidofproc by pidof
>>> command).I have no idea on which packet could be these commands. Maybe
>>> these commands are functions of Red Hat systems who are in file
>>> /etc/rc.d/init.d/functions . Someone could have a look in this file ?
>>>
>>
>> yes, pidofproc is there.
>>
>
> Now I am using generic script include on corosync-1.1.2.tar.gz .
>
> I have found that when I think that corosync is down, or isn't up and
> killall command is confused, it is not really true. The process that
> corosync is sawing is corosync init script because I have called corosync.
>  This is the confussion of  killall command to not start well or to try to
> shutdown forever.
>
> But, sometimes corosync is not starting... I have attached the daemon.log of
> one ot these times when corosync is unable to start. Could you have a look.

Well it looks like corosync dies, which prevents the cib from starting
which prevents the crmd from starting.
Core file?