Oracle 10G SGeRAC ServiceGuard config assistance please.

[isaac_loven]

Hi all,
We have created a cluster for our new RAC test system, but we have some final config issues.
We have SGeRAC Vers 11.17, and RAC 10.2.0.2.0,
and patches have been updated too.

In the cluster cmclconf.acsii file we have set "OPS_VOLUME_GROUP /dev/vgractst".
But we have found that we need to run by hand "vgchange -a s vgractst" on each node after a reboot and before the cluster is started.
Is there a recommended way of incorporating this vgchange in cluster startup ? Or should I put this VG into the /etc/lvmrc file ?

We also wish to run "crsctl start crs" after the cluster is started. Should I put this into a package with "PACKAGE_TYPE = MULTI_NODE" so that it is started on all nodes? Or should this be done by hand?

( I tried to find some docs to help with no success )

Thanks, Isaac.

[Denver Osborn · Accepted solution]

The volume group activation should be handled by the package. Your package control script should contain a VG[] array for each of your OPS volumes. There should also be a VGCHANGE variable. If you have the VG[] entry already defined in the package, then see if the VGCHANGE var is set as "vgchange -a s".

Details can be found online too, http://docs.hp.com/en/T1859-90017/ch02s08.html

As for crs, look here for a starting point.
http://docs.hp.com/en/T1859-90038/ch02s21.html

You will want to configure clusterware to start after rac and shutdown before you bring it down.

Hope this helps,
-denver

[Steven E. Protter]

Shalom,

When you say patches are updated, does that mean that all Oracle required patches were installed prior to RAC installation?

If major or Oracle required patches were added after RAC install, an Oracle relink is required.

Note also, that Oracle RAC's shared storage must be activated in shared mode. This makes it impossible to make adjustments, or use OS utilties to make storage changes without downing RAC and the database. Make sure you are happy with the LVM configuration prior to going live.

SEP

[isaac_loven]

Thanks Guys for your help. Steven, we patched the boxes before loading Oracle, so no relink is necessary - but it is good to remember for the future.
Thanks Debver for pointing me in the right direction.

The RAC Cluster is now succcessfully working.
Below is the solution, so I can help others:

The solution: =======================
run on each node once only to:
/oravl01/oracle/10.2.0/bin/crsctl disable crs ;
/oravl01/oracle/10.2.0/bin/crsctl stop crs


In pkg.control script:
VGCHANGE="vgchange -a s"
and the start commands:
/sbin/init.d/init.cssd start
/oravl01/oracle/10.2.0/bin/crsctl start crs
and stop in the reverse order.

In the pkg.conf:
PACKAGE_NAME crspkg1
PACKAGE_TYPE MULTI_NODE
AUTO_RUN YES
LOCAL_LAN_FAILOVER_ALLOWED YES
NODE_FAIL_FAST_ENABLED YES

Also Do: cmmodpkg -e -n for ALL nodes.

[isaac_loven]

Also first make the VG Sharable on all nodes:
vgchange -a n vgractst ## all nodes.
vgchange -S y -c y vgractst ## all nodes
vgchange -a s vgractst ## all nodes.
vgdisplay -v vgractst ## to view status

Deactivate before adding to Service Guard.
vgchange -a n vgractst ## all nodes