1

In previous version of Oracle RAC with Oracle clusterware, before 10.0.1 I believe, an instance was dependent on a VIP resource. Once a vip was failed or relocated, the Instance would be stopped. In later versiond, this coupling was broken, as I understand it, by costumer requests.

Since an instance is not a regular resource in the CRS cluster, and can not be configured with crs_profile, how can I create a dependency between an Instance and it's VIP?

katriel
  • 4,407
  • 22
  • 20
  • I'm wondering what is the reason behind this? Are you sure you don't misunderstand the relations between VIP and listener and between the listener and the instance? – kubanczyk Dec 02 '11 at 23:23
  • The reasoning was a service (Streams apply) that was running on an instance, and would not move to the other instance on vip failure. This would cause synchronization problems with every network problem – katriel Dec 05 '11 at 07:09
  • What was preventing the secondary listener to serve this service? – kubanczyk Dec 05 '11 at 19:01
  • I am not a DBA, but I was told it has something to do with an apply queue running in the specific instance. – katriel Dec 06 '11 at 06:25

1 Answers1

2

Solved! The trick is to re-register the instance resource. As root:

crs_stat -p ora.db.instance1.inst > /tmp/ora.db.instance1.inst.cap

Edit /tmp/ora.db.instance1.inst.cap and add the vip resource to REQUIRED_RESOURCES

Run:

crs_register ora.db.instance1.inst -d /tmp -u

the "-u", I'm guessing is update, does the magic. After that you need to re set permission (or maybe run the whole thing as "oracle" user)

crs_setperm ora.db.instance1.inst -o oracle
crs_setperm ora.db.instance1.inst -g dba
crs_setperm ora.db.instance1.inst -u pgrp:dba:rwx

That's it. Run a 

crs_stat -p crs_stat -p ora.db.instance1.inst

to make sure the changes have been saved and restart the instance

This is relevent to 10.2.0.4

katriel
  • 4,407
  • 22
  • 20