[Linux-HA] Node failure causes peer host to reboot?!?

[Andrew Beekhof]

On Thu, Apr 17, 2008 at 12:58 PM, Andrew Beekhof <beekhof at gmail.com> wrote:
>
> On Thu, Apr 17, 2008 at 12:56 PM, Andrew Beekhof <beekhof at gmail.com> wrote:
>  > On Thu, Apr 17, 2008 at 12:35 PM, Luis Motta Campos
>  >  <luismottacampos at yahoo.co.uk> wrote:
>  >  > Dejan Muhamedagic wrote:
>  >  >  > Hi,
>  >  >
>  >  > >> respawn hacluster /usr/lib64/heartbeat/ipfail
>  >  >  >
>  >  >  > ipfail doesn't work with crm. You should use pingd instead.
>  >  >
>  >  >  Well, I don't think this helps. :( I'm using the suggested (reasonable
>  >  >  for me) defaults:
>  >  >
>  >  >  respawn root /usr/lib64/heartbeat/pingd -m 100 -d 5s
>  >  >
>  >  >  (yes, I'm running CentOS x86_64).
>  >  >
>  >  >  I still have problems, but they seem to be worse, now. Before, if I
>  >  >  restarted heartbeat (/etc/init.d/heartbeat restart), any service running
>  >  >  on the machine jumped away before the restart, and heartbeat was able to
>  >  >  restart ok.
>  >  >
>  >  >  Using pingd instead of the ipfail, even this is crippled, and heartbeat
>  >  >  reboots the peer host (the one supposed to keep services running) if I
>  >  >  try to restart the heartbeat service on one of the machines.
>  >  >
>  >  >  I presume I'm doing something really stupid, but I can't understand it.
>  >  >  Please help me out. I used hb_report to fetch all I know about my
>  >  >  system, please find the report attached.
>  >  >
>  >
>  >  random question - did you install from source or packages?  where did
>  >  you get them from?
>  >
>
>  and a followup... you cant just make up values for target_role:
>
>              <nvpair name="target_role" value="Started:Master"
>  id="d54bdbb8-5d79-4d12-a95f-9b9b015176e3"/>
>
>  makes no sense.  just "Master" would be correct
>

Then there is the failed start operation... that wont be helping at all.

pengine[13743]: 2008/04/17_12:23:22 WARN: unpack_rsc_op: Processing
failed op database-filesystem_start_0 on db-sql1.ripe.net: Error

And finally, it looks like there was a crash in the pengine process.

crmd[12352]: 2008/04/17_12:23:22 WARN: Managed pengine process 13743
killed by signal 11 [SIGSEGV - Segmentation violation].
crmd[12352]: 2008/04/17_12:23:22 ERROR: Managed pengine process 13743
dumped core

can you have a look for a core file in
/var/lib/heartbeat/cores/hacluster/ and post the backtrace?