Friday Jan 18, 2013

Once Upon a Time in the Kingdom of Serv

If you're the type of person who has no time to read fairy tales, scroll to the very bottom for a link to the article.

Once upon a time there was a very happy Kingdom called Serv. It was ruled by inventors called engineers. Most of the engineers were clever, kind, and handsome. They had beautiful wives who cooked them tasty and nutritious meals.

A few of the engineers, however, had wives with big, hairy, purple moles, who sat around all day watching reruns of Bridezilla while chomping loudly on pork rinds. They never served their engineer husbands any meals and instead, screamed at them to get them another bag of pork rinds. And they hated sysadmins.

Sysadmins were the workers of the Kingdom. They were very playful, and they had big strong hands. They spent their days tossing servers back and forth to each other, or playing hacky sack.

The Kingdom was a happy place because the clever, kind, and handsome engineers had long ago invented a wonderful contraption called, as you would expect, a "server." Servers were loved throughout the Serv kingdom and all the surrounding kingdoms. They came in shiny metal boxes and had blinking lights. Best of all, they had straight edges so that sysadmins could toss them back and forth to each other. Sysadmins loved tossing servers back and forth to each other, and at lunch time it was not uncommon for several servers to be in the air at once. But when a sysadmin dropped a server, it usually broke. And when a server broke, it was called a "failure." And a failure always woke up The Boss.

The Boss was a hairy ugly giant with one eye. He did only two things. He slept. And he fired sysadmins for waking him up. Naturally, everybody preferred to keep the boss asleep. Especially sysadmins.

Polite people in the Kingdom never mentioned the word "failure" at dinner parties, not even in a whisper, lest they unwittingly awaken The Boss. But everybody knew that if sysadmins began to appear on their sofas in the middle of the night, somewhere in the Kingdom a failure had occurred.

The wives of the clever, kind, and handsome engineers begged their husbands to do something about the plight of the playful sysadmins. And so the clever, kind, and handsome engineers invented the cluster. A cluster was an enchanted cable that connected groups of servers in a magical way. When one server was dropped by a sysadmin, the cable moved that server's applications to another server so fast that nobody had time to even think of saying "failure," much less say it loud enough to wake The Boss. When the dropped server was fixed, the enchanted cable moved that server's applications back.

And so the Kingdom was full of happy sysadmins tossing servers back and forth during lunch, and sleeping in their very own beds at night.

This turn of events, of course, made the pork rind and Bridezilla wives jealous. During the commercials they screeched at their browbeaten husbands until they invented a curse to get the sysadmins fired again and back on the sofas of the beautiful wives who cooked their engineer husbands tasty and nutritious food.

It was an unspeakable curse, and polite people at dinner parties didn't dare to even whisper its name. When this curse was unleashed upon the Kingdom, all the beautiful metal servers disappeared. Except one. And inside that one server were trapped the spirits of all the other servers. The sysadmins stood around staring at it, wondering of what use their big strong hands were when the servers no longer had bodies.

One by one the sysadmins grew sad and left, and in no time at all, almost all the clever, kind, and handsome engineers had sysadmins sleeping on their sofas again.

The Kingdom was not a happy place.

Until one day, it occurred to the cleverest, kindest, and most handsome of the clever, kind, and handsome engineers to put a spell on the enchanted cable so that it could do the same thing for the spirit servers that it once did for the physical servers.

It was a wonderful invention, and the sysadmins jumped off their sofas to learn how to use it. And to keep the pork rind-chomping, Bridezilla-watching wives of the browbeaten engineers guessing, the enchanted cable could be used in two different ways:

Two Ways to Create a Cluster from Logical Domains

  • Configure logical domains within Oracle Solaris Cluster
  • Configure Oracle Solaris Cluster within Oracle VM Server for SPARC

The first approach is fairly obvious. You can put one or more applications inside each domain and create a cluster from all the domains. When a particular domain goes down, the applications running inside it get moved to a working domain. The domains are controlled individually through Oracle VM Server for SPARC, and the cluster is controlled by Oracle Solaris Cluster.

The second approach is more involved, but it provides significant benefits. It consists of setting up Oracle Solaris Cluster inside the control domain of Oracle VM Server for SPARC. When deployed this way, Oracle Solaris cluster can manage guest domains as "black boxes," which allows a site to isolate the administration of guest domains from each other. With this approach, from within Oracle Solaris Cluster you can:

  • Create guest domains
  • Live- and warm-migrate the guest domains
  • And manage individual applications like you can with the first approach

The second approach is well documented. In fact, Venkat Chennuru, a sysadmin with big strong hands who was elevated to the rank of clever, kind, and handsome engineer, took the trouble to write it down for us. You can find his article on OTN:

How to Configure a Failover Guest Domain in an Oracle Solaris Cluster

Read it, learn how to do it. Because as you know, evil never rests.

- Rick

Website Newsletter Facebook Twitter My Personal Blog

Wednesday Sep 12, 2012

Reaping the Benefits of the Image Packaging System

source

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.

Resources

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 11.2.0.2 on Oracle Solaris Zone Clusters 11.2.0.2 10 3.3
How to Deploy Oracle RAC 11.2.0.3 on Oracle Solaris Zone Clusters 11.2.0.3 10 3.3
How to Deploy Oracle RAC 11.2.0.3 on Oracle Solaris 11 Zone Clusters 11.2.0.3 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

Tuesday May 22, 2012

Cluster Fest

It's been a couple of months since we focused on Oracle Solaris Cluster. If you're a fan, we have some new content that will interest you. See below. (If you're new to Solaris Cluster, in particular how to use it in a virtual environment, see "Recent Technical Articles About Oracle Solaris Cluster," further down.)

New Technical Articles About Oracle Solaris Cluster

How to Upgrade to Oracle Solaris Cluster 4.0
If you are running Oracle Solaris Cluster 3.3 5/11 on Oracle Solaris 10 and want to upgrade to Oracle Solaris Cluster 4.0 running on Oracle Solaris 11, consider using the Oracle Solaris Cluster Geographic Edition software. It makes the job easier and keeps downtime to a minimum. Tim Read wrote this 8-part article to show you how. Contents are:

How to Deploy Oracle RAC on Zone Clusters
This one is very cool. Oracle Solaris Cluster lets you create clusters of Solaris zones. That gives you high availability. You also get high availability from Oracle Real Application Clusters (RAC). So why would you install RAC on zone clusters? Because you can implement a multi-tiered database environment that isolates database tiers and administrative domains from each other, while taking advantage of centralized (and simpler) cluster administration. This article explains how to do it.

Recent Technical Resources About Oracle Solaris Cluster

Blog: How to Survive the End of the World - Part I
Provides a simple example of a two-node cluster, and provides resources to help you create one.

How to Survive the End of the World - Part II
Changes the 2-node example above into a failover cluster, and provides resources to help you create one.

As always you can find the latest technical resources to help you evaluate, test, and deploy Oracle Solaris Cluster on OTN's Cluster Resources for Sysadmins and Developers

- 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
--------------------------------------
footballDitka
cheerleadersDallas
beerBud
bratsMilwaukee
--------------------------------------
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
--------------------------------------
footballfootball-zone
cheerleaderscheerleader-zone
beerbeer-zone
bratsbrats-zone
--------------------------------------
civilizationcivilization zones

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

Zone on EarthFailover ZoneDistant Planet
---------------------------------------------------------
football-zonefootball-failover-zoneDitka
cheerleaderscheerleaders-failover-zoneDallas
beerbeer-failover-zoneBud
bratsbrats-failover-zoneMilwaukee
---------------------------------------------------------
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
Website
Newsletter
Facebook
Twitter

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:

football
+cheerleaders
+beer
+brats
-------------------
civilization

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

Building BlockBackup Planet
-------------------
footballDitka
cheerleadersDallas
beerBud
bratsMilwaukee
--------------------------------------
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

Website
Newsletter
Facebook
Twitter

Monday Jul 18, 2011

Women Soccer Players Don't Flop

I'm now a rabid fan of women's soccer. Men's soccer? Phht! Men flop. Really. They could recruit Vlade Divac straight out of the NBA without letting him even see what a soccer ball looks like, and he'd be a superstar before mid-season.

Sunday's match between Japan and the US was one of the best matches I've ever seen, in large part because the women played clean soccer. And because they exhibited none of the honor-sapping flopping that is prevalent in the men's game.

What's this got to do with Oracle Solaris Cluster? Think about it. You can hit the Japanese women with a veritable tsunami of shots on goal, you can overwhelm them with skill, speed, power, and intensity. You can jump over them, power through them, score debilitating goals on them. More than once. And you can run them into the ground for 120 minutes. But they never go down. They never give up. They keep fighting. And they keep coming at you.

Wouldn't you want that on your side?

We just released version 3.3 5/11. Here are some resources to help you become familiar with it.

Resources for Oracle Solaris Cluster 3.3

By the way, that picture is not from the USA-Japan match, but from nj.com.

- Rick
System Admin and Developer Community of OTN
OTN Garage Blog
OTN Garage on Facebook
OTN Garage on Twitter

About

Contributors:
Rick Ramsey
Kemer Thomson
and members of the OTN community

Search

Archives
« April 2014
SunMonTueWedThuFriSat
  
1
2
3
4
5
6
7
8
9
10
12
13
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
   
       
Today
Blogs We Like