How to make the node reboot automatically

[Wang,MinJie]

Hi all
I've configured an Oracle Cluster
Now there is a question coming
When I fail over an Oracle Package from node A to node B node A doesn't reboot automatically
How to configure it
Thanks in advance

[A. Clay Stephenson · Accepted solution]

Your question is far from clear. Are you trying to make the package shift back over to the original node or are you trying to get the original node to reboot after you deliberately crashed it after doing something like yanking power.

You may have done something which triggered a TOC to avoid a split-brain condition because the node split into exactly equal parts and could not get a quorum lock. Normally, rebooting has nothing to do with SG and everything to do with how the boot is configured in stable storage. If the box did do a TOC then you definitely do not want it to reboot without human intervention. In fact, I never allow packages to failback automatically because the number of reasons for a package failover is legion and I would much rather evaluate (and possibly correct) the condition before I ever allow t5he package to run again on the original node.

[Wang,MinJie]

Thank you for your advice
I'm a freshman on MC
Now I want to make sure my MC is Ok by some test.
I've killed every important process of Oracle (such as chpt smon pmon) and node B will substitute node A immediately
Does that mean my MC for oracle is successful?
If not,what else should I test?
Any advice?

[Safarali]

hi wang if you want to test your MC is working either you have to reboot your primary system or pull the lan cable from primary then test it is moving to another node

Regards
Safar

[Wang,MinJie]

Hi Safarali
I've tested my cluster according to your advice
When I pull the lan cabel of Node A
Node B will substitute
But when I reboot Node A which by the way is running the package
A system crash will occur on Node B and Node B will reboot too

The question is:
what parameter is responsible for that if node A reboot node B will substitute node A?
Thank you

[likid0]

Hy, it seems like a cluster reformation issue.

is your cluster lock disk or quorum server correctly configured?

You can check /var/adm/syslog/syslog.log

to get a better idea of what is exactly happening when your nodeB tocs itself.

[Wang,MinJie]

Hey Orange
I've checked the log it said:
WARNING: Cluster lock on disk /dev/dsk/c3t2d0
is missing
cmcld[3514]: Until it is fixed, a single failure could
cmcld[3514]: cause all nodes in the cluster to crash


"/dev/dsk/c3t2d0" is my lock pv of cluster
by the way

So what should I do?

[Mridul Shrivastava]

Yuo have to halt the cluster once before that you need to configure the lock disk in cluster.ascii file.

Then halt the cluster and activate that vg on one node. then run cmcheckconf followed by cmruncl if there is no error.
sequence should as follows.. this is an example:

# cmviewconf | grep -i -e "node name" -e lock
# cd /etc/cmcluster
# cmgetconf cluster.ascii
Edit the cluster.ascii file in order to include the new lock disks on it.
# cmcheckconf -C cluster.ascii
# cmhaltcl -f
# cmapplyconf -f -C cluster.ascii
# cmruncl

don't forget to take vgcfgbackup of the lock vg then deactivate it. As once this is done that VG shouldn't be active on any node.