Boot Error

[Zulfikri]

Hi all,

I have to boxes running service guard. BOX A and BOX B.
Last nite, BOX A rebooted for no reason. No logs whatsoever telling the error.

But BOX A now are booting up as BOX B. BOX B already live and running all my application.

Is there a problem here? How do I get back BOX A?

thanks in advance

Zulfikri

[Steffi Jones_1 · Accepted solution]

Hello,

there are several places where you can look for a hint:

- var/adm/syslog/OLDsyslog and see what the cmcld complains about
- /var/adm/crash for a crashfile
- /etc/shutdownlog for a shutdown message

If you don't find anything at all check if you have power problems.

I don't think that I understood the problem correctly that box a is coming up as box b. How do you see this?

Maybe you can feed us some more information,

Steffi JOnes

[Zulfikri]

Hi all,

I've checked /var/adm/crash and /etc/shutdownlog

In shutdownlog it has reboot after panic, isr.ior = 0,9227ffff.c00000000'

in crash it contained bounds,crash.0 crash.1(on the time and date it rebooted)
crash.1 has INDEX and some crash*.gz files.

Normal operation is that we have both BOX A and BOX B running. BOX A has all the softwares nodes up and if it fails, it will be transfered to BOXB

Right now BOX B is ok and running all the nodes, but BOX A is booting up as BOXB with all the ip addresses and hostname. It really puzzling me.

Please help

Thanks in advance

Zulfikri

[James R. Ferguson]

Hi:

In addition to Steffi's suggestions, take a look at the package control logs (e.g. /etc/cmcluster//control.sh.log)
for clues, too.

...JRF...

[James R. Ferguson]

Hi again:

Make sure that your /etc/rc.config.d/netconf contains the correct hostnames and ipaddresses on both nodes.

...JRF...

[Carsten Krege]

One more idea in addition to what James already suggested.

Is it possible that the boot disks are on a shared bus and that nodeA boots from the wrong disk? In this case you should check offline(!!) whether the boot pathes in the boot menu point to the same disk on both nodeA and nodeB.

Be careful: Data corruption is very likely to occur when both systems boot from the same disk and are up at the same point of time!!

This is one reason why it is not recommended to have the boot disks on a shared bus. In fact the only supported config is that nodeA has its primary root disk connected to one shared bus and nodeB has its primary root disk to another shared bus (due to IODC restrictions though).

In former times this has sometimes be seen with the now obsolete Switchover/UX software. Part of the Switchover concept was to have the boot disks on the shared bus and to let the standby node boot from the primary's boot disk when the it had problems. Too bad when the primary did not go down before!

Carsten

[Steffi Jones_1]

Hello,

one more thing...
If you have a isr,ior for the time the box went down you should look into getting the box checked out for possible hw errors.

Steffi Jones

[Bill McNAMARA_1]

You want to verify that your boot paths are correct. If BOXB is booting of of BOXAs boot disk, this could explain your problem.
lvlnboot -v
and from BCH, PA to display boot paths.
A careful mapping of your system from ioscan -fnk and strings /etc/lvmtab could indicate lvm misconfiguration.

Bill

[CHRIS_ANORUO]

To check the boot path use:

echo boot_string/S | adb -k /stand/vmunix /dev/mem

As has been suggested, check /etc/rc.config.d/netconf files for ip address and hostnames.