Oracle HAIP

Starting with 11.2.0.2 oracle introduced highly available IP (HAIP) addresses. This allows the oracle clusterware to provide redunadant interconnect usage without the need for bonding at the network interface level. More information on HAIP can be found here for Solaris.

Lets see it in action. We have defined network interfaces across two vlans and the interfaces are vnet2 and vnet3. Our interface setup prior to install 12C GI looks like this:

Now I wont go into details on the GI install but there are a couple of ways you can add these interfaces. Using the GUI select 2 private interfaces or use a response file and do a silent install which is the method I’ve been favouring. Response file config parameter show below for my test lab server. Please note the setting below does not have the second passive IPMP public interface due to an Oracle bug as I’ve disucess in this blog post.

After a successful install of GI you should see the GI stack do some plumbing on the interfaces.

You can see grid has plumbed in some 169.254 addresses. These are described in the metalink doc Grid Infrastructure Redundant Interconnect and ora.cluster_interconnect.haip (Doc ID 1210883.1)
but are basically reserved local link addresses which should not be used for any other purpose and described in RFC-3927.

Now lets down an interface and see what happens.

Looking at some of the logs we see the following.

ohasd.log

crsd.log

If we look at our interfaces with ifconfig we can see GI has failed it over and plumbed it into vnet3

All resources are still online.

Now if we bring the interface back up we see the following.

You can see the local link address has failed back to the vnet2 interface. If you check the orarootagent_root.log you can see a little bit more detail on what is going on.

Thats a very basic look at HAIP functionaliy on Solaris with a failover test.

Leave a Reply

Your email address will not be published. Required fields are marked *

CAPTCHA * Time limit is exhausted. Please reload CAPTCHA.