Wednesday Sep 12, 2012

Reaping the Benefits of the Image Packaging System


One of the promises made about Oracle Solaris 11 was easier installation. Remember?

Do you also remember how involved installing Oracle Solaris Cluster used to be? It was so involved, in fact, that we (when we were Sun Microsystems) wouldn't even let you do it yourself.

How times have changed.

New - How to Automate The Installation of Oracle Solaris Cluster 4.0

Thanks to the new image packaging architecture in Oracle Solaris 11, you can now automate the installation of Oracle Solaris Cluster 4.0. Why is that such a big deal? As Lucia Lai explains it:

"Without the AI, you would have to manually install the cluster components on the cluster nodes, and then run the scinstall tool to add the nodes to the cluster. If, instead, you use the AI, both the Oracle Solaris 11 and the Oracle Solaris Cluster 4.0 packages are installed onto the cluster nodes directly from Image Packaging System (IPS) repositories, and the nodes are booted into a new cluster with minimum user intervention."

Lucia goes on to explain how to set up and configure the AI server, how to plan your cluster configuration for the automated installation, how to use the scinstall utility, how to set up the DHCP server, and more. A thorough, well-written article.

- Rick

Website Newsletter Facebook Twitter

Friday Aug 03, 2012

My Oracle RAC and Oracle Solaris Cluster Cheet Sheet

This gets complicated, so stop watching motoGP crash compilation videos for a sec.

We have Oracle Real Application Clusters (RAC). RAC lets you deploy a single Oracle Database across different servers. If the server in your Des Moines data center gets picked up by a tornado that hates you and dropped off in East Texas, the other servers pick up the load, and the database continues to operate without interruption. That's easy to understand.

We also have Oracle Solaris Cluster. It lets you deploy the Oracle Solaris operating system across different servers. If the server in your Barbados data center gets washed away by a hurricane that hates you and dropped off in West Africa, the other servers pick up the load, and the operating system continues to operate without interruption. A good quote:

White Paper: Extending Oracle Solaris for Business Continuity
"Oracle Solaris Cluster offers comprehensive and robust capabilities for keeping your business IT, including those running Oracle Database and Applications, up and running in the face of nearly every conceivable situation."

That's easy to understand, as well.

So why would somebody complicate our sysadmin lives by suggesting we install Oracle RAC on Oracle Solaris Cluster? What would that be, highly-available high availability?

Turns out that's not what they're suggesting. They're suggesting we install Oracle RAC not on Solaris Clusters, but on zone clusters. What's a zone cluster, you ask?

A zone cluster is a cluster created from Solaris zones that are physically located on different servers. That's similar to a regular cluster, but it uses zones instead of entire OS instances. Don't confuse a zone cluster with a failover cluster. Instead, read this white paper:

White Paper: Zone Clusters: How to Deploy Virtual Clusters and Why
This paper introduces the zone cluster, a virtual cluster in which an Oracle Solaris Zone is configured as a virtual node. The zone cluster supports the consolidation of multiple cluster applications on a single cluster.

That's all very interesting, but what about our original question:

Why would someone want to complicate our sysadmin lives by suggesting we install Oracle RAC on a zone cluster?

Turns out there two good reasons:

  • It's a better high-availability solution for a multi-tier application environment
  • It lets you isolate your database development, test, and deployment environments from each other.

How the Oracle RAC/Zone Cluster Combo Is Better For Multi-Tier Applications

Let's say that you are using your Oracle database as one tier in two different application environments. The first one is an HR application, the one second is an e-business suite. Both access the same database. Well, Oracle RAC would give you the high-availability for that database. But the applications would not be highly available. However, if you installed the database with Oracle RAC inside one zone cluster, and each application inside its own zone cluster, you'd make both application environments highly avaiable. And, if you limit the administrative privileges for each zone cluster, you'd get administrative isolation, as well.

How the Oracle RAC/Zone Cluster Combo Is Safer for Deployment

You've probably heard by now about Knight Capital Group's trading glitch that dropped the company's value by 50% in one day. I don't know exactly what happened, but I wonder if they didn't deploy either their development or their test environment instead of the one that was ready for prime time.

I suppose it's a sysadmin's duty to learn from another sysadmin's misfortune. So, if you divide your zone clusters into development, test, and deployment environments, you might have a better shot at avoiding a similar catastrophe. For example, install Oracle RAC with an Oracle DB into your development zone cluster, and keep it isolated from your test and deployment zone clusters. One sysadmin controls the development cluster. Another the test cluster. And the biggest, baddest sysadmin controls the deployment cluster. When the development environment is ready for testing, the test admin must OK the migration. That goes double for the deployment environment. And all the while, each environment remains highly available.


Turns out that Oracle and the portion of Oracle that was once Sun Microsystems have been collaborating on Oracle RAC/Solaris Cluster solutions for a long time. Customers like this approach so much that we just published three articles explaining how to do it. Each article covers a different version of the software:

Article RAC Version Solaris Version Cluster Version
How to Deploy Oracle RAC on Oracle Solaris Zone Clusters 10 3.3
How to Deploy Oracle RAC on Oracle Solaris Zone Clusters 10 3.3
How to Deploy Oracle RAC on Oracle Solaris 11 Zone Clusters 11 4.0

And if you want more, we also have a page full of links to all our Solaris Cluster how-to articles and background white papers:

Where to find everything Solaris Cluster-related

Don't be the sysadmin who bankrupts your company in one day. Get educated.

- Rick

Website Newsletter Facebook Twitter

Monday Jan 23, 2012

How to Survive the End of the World - Part II

In Part I of our Survival Guide for Civilization, I explained how to save civilization by identifying four distant planets that had the essential building blocks of civilization and combining them into a 5-node cluster with Earth:

Building BlockDistant Planet
civilizationbackup civilization

As mentioned in Part I, the resulting five-node cluster was actually more than what we wanted. Five distant planets! We'd rather not deal with the overhead of managing five distant planets. We prefer to keep managing just one planet, but make sure that can keep civilization humming. Turns out that we can accomplish that through the magic of virtualization. As you might expect, it's called a virtual cluster. (Really techie people call it a failover zone cluster.)

First, we create one zone on Earth for each building block:

Building BlockZone on Earth
civilizationcivilization zones

Then we create one failover zone on each distant planet for each zone on Earth:

Zone on EarthFailover ZoneDistant Planet
zone civilizationfailover zone civilization

In this way, each failover zone on its distant planet backs up one original zone on Earth. It's a great way to save civilization with much less overhead.

As it turns out, not only do we have an article that shows you how to create a cluster with Solaris Cluster 4.0, but we have one that shows you how to create a failover cluster, too:

How to Create A Failover Zone Cluster

Give it a try. It never hurts to be prepared.

- Rick

Thursday Dec 15, 2011

How to Survive the End of the World - Part I

If you've been paying attention you'll probably agree that Earth will be destroyed any day, now.

That used to concern me.

But the more I understand clustering, the more I realize we can simply reconstitute civilization from individual slices of other planets in the Universe. The first thing we need to do is identify the building blocks of an advanced civilization. That should be relatively simple:


Next, find planets that had excellent examples of each building block:

Building BlockBackup Planet
civilizationbackup civilization

Those four planets plus Earth would be easy enough to arrange into a high-availability cluster if we downloaded and installed Oracle Solaris 11 and Oracle Solaris Cluster 4.0 on each planet, including Earth.

With Solaris Cluster 4.0, we could create a nice five-node cluster. Not only would the cluster provide the disaster recovery we're looking for, but it would actually help us create an elastic cloud of sorts, in which we could, for instance, tap into the beer of planet Bud during the Super Bowl or other times of dire need. See What's New to read about other cool things you can do with Solaris Cluster 4.0.

Creating a five-node cluster can get a bit tricky, but you can build up your skills by creating a smaller one, using the instructions in this OTN article:

How to Install and Configure a Two Node Cluster

Once you have the two-node setup figured out, you can move to the five-node setup. But the resulting five-node cluster is actually more than what we want, isn't it? It's a cluster of five entire planets, when what we're looking for is a slice of each planet. In an upcoming blog I'll summarize how to create a cluster from the slices of those individual planets. That's called a virtual cluster or a zone cluster, and it's very cool.

- Rick


Rick Ramsey
Kemer Thomson
and members of the OTN community


« April 2014
Blogs We Like