Configuring IPMP for private cluster Interconnects

It is advisable to first have a look at the post configuring IPMP on Solaris with oracle RAC before going through this post.

With IPMP configured for private cluster interconnect extra care needs to be taken else the redundancy and high availability can inadvertently result into cause of bigger problem.I have faced this problem and hence i am sharing with you the solution of this problem

When IPMP is configured for private NIC card when running VIPCA you will select all the  NIC cards during the installation and which is also correct and required during installation.But with this configuration OCR will store the information about Interfaces and each time when the cluster is started the OCR is checked and the health of each of the NIC card is checked .If all the NIC cards are proper only then cluster will allow ASM to and hence the database to start.But Failure of any of the NIC card (primary or even “STANDBY”) will prevent ASM to start and hence will also prevent ins tance to start from that node in which any of the NIC card is failed.

In my case the standby NIC card failed which should not have caused any problem because physical IP was pointing to IP of primary NIC which was alive  but OCR  conatined the information of standby NIC also which “FAILED” and hence it did not allow ASM instance to start giving ora-27504 IPC OSD context error and stating that private interconnect has failed.So the the instance on that node on which standby NIC card failed was not allowed to start hence redundancy resulted into cause of problem.

To avoid such error we need to eliminate the nic information from OCR and set cluster_interconnects parameter in init.ora of both ASM and database instance.

Whether  NIC information is contained in the OCR can be determined from oifcfg utility

bash-3.00$  oifcfg getif
hme0  192.168.50.0  global  public
hme1  192.168.50.0  global  public
qfe0  172.16.0.0  global  cluster_interconnect
qfe1  172.16.0.0  global  cluster_interconnect
 

As shown above the output of OIFCFG contains information of both NIC qfe0 and qfe1.

then we need to delete the information of both NIC’s from OCR using oifcfg delif utility

oifcfg delif -global qfe0

oifcfg delif -global qfe1

bash-3.00$   oifcfg getif
hme0  192.168.50.0  global  public
hme1  192.168.50.0  global  public

Now a parameter called cluster_interconnects needs to be set in init.ora(or spfile) of both ASM and database instance so that it points to physical IP which will always be alive irrespective of failure of any NIC card.

 cluster_interconnects=172.16.50.6

With this modifications when the cluster is up OCR will not check for the health of NIC card and when the instance(ASM or Database) is up it will try to look for whether the IP provided by cluster_interconnect is alive and if is alive it will allow the instance to be up.

For reference u can look at metalink Doc id 368464.1 ( How to Setup IPMP as Cluster Interconnect)

If u have faced any such problem and have solved it write your comments we can share our work.

Thanks and Regards

 Parikshit