[Linux-HA] heartbeat 2.0.0, watchdog, RHEL 4

[Poyner, Brandon]

> Thanks for the information.  Actually, I thought that code 
> was no longer useful.  I thought it would autoload the 
> softdog just from opening the device.

RHEL uses udev by default, and by default /dev/watchdog doesn't exist at
boot time.  It kernel does attempt to load the module if /dev/watchdog
exists with the correct major/minor and you open it.  The problem in the
init script is that it attempts an insmod which fails, and then it tries
to find the minor in /proc/misc which it can't find, and then tries to
mknod the device but due the previous error doesn't send mknod enough
parameters.

> This would be because for some reason heartbeat exited 
> prematurely.  I would expect to find a core file somewhere...

You are correct.  Do you have any gdb commands you'd like me to run?
Heartbeat 2.0.0 is no longer installed on this machine but I do have it
on installed on a test box.

-rw-------  1 root root 41717760 Aug 17 11:50
/var/lib/heartbeat/cores/root/core.2177
-rw-------  1 root root 40771584 Aug 17 12:08
/var/lib/heartbeat/cores/root/core.2187

> I have no idea why it can't write to media pipe 1.
> 
> I think that's the unicast child process...  I have no idea 
> why you're sending bcast and ucast to the same ethernet 
> adapter...  Can you help me with this?

Ignorance on my part, I was pretty sure that wasn't the right setup but
I've been running like that for a long time under heartbeat 1.0.3.  Are
you saying I should pick one or the other?  This is a dedicated ethernet
card for the heartbeat.

> What is at address 192.168.9.1?  I assume it's the other node in the 
> cluster?

Correct.

Brandon Poyner
Network Engineer III
CCAC - College Office
412-237-3086