I've been working on a proof of concept for OBIEE 11g. Part of that, naturally, includes building out the system.

I'm a fan of rebuilding things over and over and over. Yeah, it's time consuming, but I end up learning a lot...like what can go wrong.

Which brings me to today.

OS: OEL 5 64 bit
Database: Oracle 11gR2
OBIEE 11.1.1.5

Installed the database, created the repository using the RCU and installed the OBIEE software; when it gets to the configuration steps, it fails. I think it was step 12 or 13, (re)starting AdminServer.

Retry.

Fail.

Continue.

Failed again a couple of steps down. opmn something or another.

Log files:

opmnctl start: failed.
Can not resolve obiee11g for interface any
oracle.as.provisioning.exception.ASProvisioningException
 at oracle.as.provisioning.engine.Config.executeConfigWorkflow_WLS(Config.java:872)
 at oracle.as.install.bi.biconfig.standard.StandardWorkFlowExecutor.executeHelper(StandardWorkFlowExecutor.java:31)
 at oracle.as.install.bi.biconfig.standard.InstanceProvisioningTask.doExecute(InstanceProvisioningTask.java:81)
...snip...
Caused by: oracle.as.provisioning.engine.CfgWorkflowException
 at oracle.as.provisioning.engine.Engine.processEventResponse(Engine.java:596)
 at oracle.as.provisioning.fmwadmin.ASInstanceProv.createInstance(ASInstanceProv.java:178)
...snip...
 ... 14 more
Caused by: oracle.as.provisioning.util.ConfigException: 
Error creating ASInstance instance1.
Cause: 
An internal operation has failed: Error in starting opmn server
Operation aborted because of a system call failure or internal error
Action: 
See logs for more details.
 at oracle.as.provisioning.util.ConfigException.createConfigException(ConfigException.java:123)
 at oracle.as.provisioning.fmwadmin.ASInstanceProv._createInstance(ASInstanceProv.java:317)
 at oracle.as.provisioning.fmwadmin.ASInstanceProv.createInstance(ASInstanceProv.java:166)
 ... 18 more
Caused by: oracle.as.management.opmn.optic.OpticException: Error in starting opmn server
Operation aborted because of a system call failure or internal error
 at oracle.as.management.opmn.optic.OpmnAdmin.executeCommand(OpmnAdmin.java:310)
 at oracle.as.management.opmn.optic.OpmnAdmin.startOpmnServer(OpmnAdmin.java:87)
 at oracle.as.provisioning.fmwadmin.ASInstanceProv._createInstance(ASInstanceProv.java:254)
 ... 19 more

Awesome.

Can not resolve obiee11g for interface any

I destroyed the OBIEE deployment, reinstalled the repository, tried again. Fail.

Finally gave up and started from scratch.

Opened up the documentation (finally) to see if I missed any pre-installation configuration steps. Nope.


Did I do anything unusual in the initial setup?

No.

Wait.

I changed the hostname.

Using the Google Machine, I queried for Can not resolve obiee11g for interface any which took me to this OTN post.


Hmmm...

I changed the hostname. I didn't like "new-host-4". Really?

I updated my hosts file to match what was suggested:

# Do not remove the following line, or various programs
# that require network functionality will fail.
127.0.0.1       localhost
127.0.0.1       obiee11g
::1             localhost6
::1             obiee11g

Reran everything and the configuration completed without error.

Joy.

Maybe I should have followed the instructions for Installing on DHCP Hosts...

A recent discussion at the OBIEE Enterprise Methodology Group reminded me of something...

A few weeks back I tried out the YSlow extension from Yahoo against OBIEE 11g.

I couldn't figure out how to print or export the results until this morning when I clicked on the Tweet button. That gave me a URL.

For my test this morning, I used the SampleApp Build 10502 General Index (which rocks by the way, tons of great stuff and examples the team has created):



Overall Grade, D.

This page either got an A or a F. The F's were in the following categories:

- Make fewer HTTP requests
- Use a Content Delivery Network (CDN) (no idea)
- Add expires headers
- Compress components with gzip
- Put JavaScript at bottom
- Reduce the number of DOM elements
- User cookie-free domains



You can find the page here.

Interesting stuff.

Google also has an extension called PageSpeed which you can find here.

From The Twitter Machine today:



Which takes you here.

I had the pleasure of meeting Mr. Millsap in person last year at OOW. Since then, I've seen him present at KScope and more recently at our local user group.

I can't tell you how much I haven't processed yet.

Anyway, he had mentioned to me that he was considering this one day Masterclass event, but hadn't figured out all the details. He has now.

So it looks like he only has an event set up for Dallas/Ft. Worth, I want him to come back and visit Florida. We're often left out of these types of events (everyone neglects America's...ok, I won't go there).

Why?
I have an obsession about proving things...I'm tired of getting beat down by people who say "prove it."

With that in mind, I'm starting a petition of sorts, to lure Mr. Millsap into coming back to Florida.

If you're interested, email me to let me know you're interested in having this event down here in Florida. If enough people are interested, I'll bug Mr. Millsap until he agrees to come back.

Update
When Mr. Millsap was in town, after he presented, he was showing us some material from this class he was thinking about.

If you read Jake's musings over at the AppsLab, you'll know he (and Rich and Anthony) have gone to Google IO for the past 2 years. This year they came away with, among other things, a Samsung Galaxy Tab 10.1 and a Chromebook (which shipped very recently).

Mr. Millsap is doing something similar. Not only do you get to learn all about trace data, you get the tools to do so:

So sign up for the DFW class, or let me know so we can bring Mr. Millsap back to Florida.

Version: 11.1.1.5
OS: OEL 5

First of all, it's a Session variable, not a repository variable. Whatever.

So I had created a PIPELINED function to serve as the source of a table. From the analysis side, we wanted to use a presentation variable to set a session variable that was referenced in the table source.

Here's the SQL for the Stored Proc call:

SELECT *
FROM TABLE( get_key_value_pairs( 4 ) )

I hardcoded 4 in there just to see how it worked. BTW, Gerard Nico describes the process here.

I then created a session variable along with the init block.

SELECT 5 FROM dual

which would populate the P_LIMIT session variable. Nothing terribly fancy.

I then ran a simple report based off of the new table source (with the hardcoded value):



Cool, now let's swap out the hardcoded value with the reference to the session variable.

Since I don't do a lot of front-end stuff, this post at Siebel Essentials has been my bible.

I locate what I am trying to do in the matrix:

VALUEOF("NQ_SESSION.P_LIMIT")

Like a big jerk, I have to add spaces to it, you know, for readability. My SQL looks like this:

SELECT *
FROM TABLE( get_key_value_pairs( VALUEOF( "NQ_SESSION.P_LIMIT" ) ) )

OK, let's try without the quotes ("):

SELECT *
FROM TABLE( get_key_value_pairs( VALUEOF( NQ_SESSION.P_LIMIT ) ) )

Same thing.



Let's gid rid of the spaces (I've had issues with spaces before, but I can't seem to find the post).

SELECT *
FROM TABLE( get_key_value_pairs( VALUEOF(NQ_SESSION.P_LIMIT) ) )

Voilá!

Fun. I have noticed that certain areas of the RPD treat spaces slightly differently. If I ever find the other post, I'll link it up here.

Lesson? Be careful with your space.

6.56. That's about how old the little monster is now. She'll be 7 in December. Wow.

Just about a year ago, she had been admitted to the ICU at St. Joseph's Children Hospital. She spent 7 days or so on a ventilator.

Fortunately for her (and us), she's been strong as an ox over the past year. No visits (knock on wood) since then.

There have been a few moments, oh shit moments; she's coughing or congested and we think "oh no..." but nothing serious enough to warrant a visit to the "Man in White."

Where do we stand?

Kate is still not talking, but is making progress, however slow it might be.

I learned today that she knows how to sign "sorry" (as she apparently stabbed her teacher with a pencil today and also smacked one of her friends).

She's still not able to walk up the stairs. That means I am safe, unless she decides to crawl. If I'm on a call around 3, I warn the other participants that I'm about to be attacked by a little monster.

Her hair is super long now and I try to put it in a pony tail every day. She doesn't care much for that, but I try.

She'll be in the 1st grade this year, which boggles the mind. First, she's old enough to be in first grade? Second, WhyTF is she in first grade?

Of course she still lights up a room. As Dad, I am her universe...no pressure there huh? Makes me happy though, tough to complain about that.

Someday I'll get around to starting her charity; currently, if you buy John Piwowar's eBook, that money goes to her. T-Shirt money does too. Let's not forget the tons of money I make from Google for hosting ads!

For now though, I just enjoy.

I can't remember the first time I was introduced to the term, I'm guessing it was via The Waterboy and Mama Boucher:

There's a thread on Google+ related to my last post, Managing Database Entries (tnsnames.ora) that rekindled the word for me.

Today I posted something to the effect on Twitter. I've decided to start adding things I would consider The Devil, in Mama Boucher's terms, here.

Cary Millsap is The Devil. Why? A year or 2 ago he posted a link on Twitter about why we don't need to double-space anymore after a period. Each and every time I go to double-tap the space bar, I think of that post and curse Cary appropriately.

Triggers are probably The Devil. Rarely have I seen them implemented in a halfway decent manner. Usually, they're used as some work-around because someone was too lazy to update their PL/SQL...or just couldn't figure out a way to accomplish their goal without them (say, like removing direct INSERT/UPDATE/DELETE privileges on the source table...?).

From that last post, tnsnames.ora can be The Devil.

I would say that commas in front of the line are The Devil.

My friend Jason Baer is The Devil. Go to 2:40 in:



That's all I have for now.

What do you consider The Devil?

Update
Me. I can think of at least 47 people who believe I am The Devil. Man...how could I forget something like that.

For the past 18 months or so I've been arguing that if you would just manage your tnsnames.ora file, things would be much easier. I'm talking about through your various SDLC environments, DEV --> QA --> PROD.

This is true for tools like OBIEE.

In OBIEE, you can use the OCI client and specify your tnsname entry for a particular connection. Of course you'll still have to change the password for the connection, but you'll never be at risk of accidentally connecting to the wrong database.

Guess what? My worst case scenario occurred.

Here's what I have suggested; use a generic name for your connection. Let's say TESTING. How it works right now is we have one for each environment, TESTING_DEV, TESTING_QA and TESTING_PROD.

Each server has a tnsnames.ora file with every single connection.

For OBIEE, that means not only changing the password for each environment, but changing the DNS (TNS) entry as well.

Only bad things can come from this.

(I realize there are much more sophisticated ways of managing this, OID (I believe) for instance, but that's outside the scope).

So what happened?

Last week I built out an Informatica PowerCenter server which connected to the Dev database. Following that, I wrote up the instructions, including the particulars of my installation.

The next person up the chain, installing the QA software, read it literally.

Guess what happened?

The Dev server got borked because they used the Dev connection information and all the configuration stuff (technical term) got messed up.

Guess what else?

I get to rebuild the dev machine.

I would contend that the QA server should only have an entry for the QA database...that way this type of thing would never occur. If we had used a generic name for the database, say TESTING, I wouldn't be working tonight.

Something to think about when you end up managing, not only multiple servers, but multiple "platforms" as well.

Alternate Title: Cary Millsap is Awesomesauce.

Last night the Suncoast Oracle Users' Group hosted Cary Millsap of Method-R Corporation. You may have heard about Mr. Millsap, he's written a few books, a few papers, done a couple of presentations...

Thinking Clearly About Performance
Find his paper here.

While the paper is good, I don't think it compares to the presentation.

This was my first second Millsap presentation I've gotten to see this year, the last was in June in Long Beach on My Case for Agile.

We were treated to 2+ hours ending a little after 9. There wasn't a dry eye in the house. Oh wait, that was something else. It was a packed house, approximately 50 attendees including Enrique Aviles (aka "evil Les" or @eaviles94) and his friend Eric, who drove down from Orlando.



One of my (many) takeaways was finally understanding what the prepare, execute and fetch were.

Here's the sequence diagram...



Here's the profile...



I had never thought of it in the context of what something like java does. Once I did, easy. Here's some pseudo-code:

ps = prepareStatement( "SELECT table_name FROM user_tables" );
rs = executeStatement( ps );

OK, not the best example, but it helped me finally make the connection.

It's All About the Skew
So you have CPU, Network, IO and Something Else. Is it divided evenly among the 4 components (talking about performance here)? 25% CPU. 25% Network. 25% IO. 25% Something Else. No, typically it's not. When you see the spike, you know that's where you start to look (is that Axiomatic?). I can't do it justice (pun intended)...so read the paper.

The beauty of this approach though is that it's not just confined to database performance, it's software performance in general. I think the "software" people get it more than we database people do too. You have to instrument your code if you are ever going to be good at Performance Tuning.

If you ever have the opportunity to see Cary speak on this topic (really, on any topic, he's incredibly good a breaking down a complex topic and making it digestable by mortals), don't hesitate, go see it.

The Future
Cary's got some plans for the near future too, so you may just get that opportunity sooner than you think. I won't spoil it, but I will announce it after he goes public.

A little over a month ago, I saw this tweet from Neil Kodner (@neilkod).



That links to this set of data on github (using gist?).

The data gets interesting the further down you go.

Annoying, yes, but no big deal.

I was then off to KScope for a week of fun.

This weekend, I decided to open up Google Reader for the first time in, well, a long time. Scanning through my Oracle folder, I found the little gem.

Data Science Fun with the OOW Mix Session Voting Data

That was written by Greg Rahn (@gregrahn) on June 23rd, 6 days after Neil's post.

On July 1st, the winners were announced on the Oracle Mix blog.

A little tidbit from Greg's post that Bex (see below) quoted as well:

-- number of users who voted for exactly one author
+---------------------------+
| users_voting_for_1_author |
+---------------------------+
|                       828 |
+---------------------------+

-- number of voters who voted for every session by a given author
-- and total # of votes per voter is the same # as sessions by an author
+-------------------------------------------------+
| users_who_voted_for_every_session_of_an_author |
+-------------------------------------------------+
|                                             826 |
+-------------------------------------------------+

Yikes.

I think it's safe to say that Greg's analysis predicted the outcome.

It's also...unfortunate.

Which brings me to this post by Brian 'Bex' Huff today, Has Oracle MIX "Suggest-A-Session" Jumped the Shark???

I'm thankful that Bex wrote this up because, quite honestly, I was scared...to rock the boat I guess.

This conversation needs to take place. I don't believe anything was done illegally, but I think it broke the "spirit" of the rules that were set up.

I'm all for using Twitter, your blog, whatever ("whatever" gets us into trouble) to get people to vote for your stuff; but this seemed to go to far.

Bex has some suggestions for changing the rules over there, check out his piece and chime in.

We (SOUG) have the fortunate pleasure of hosting Mr. Cary Millsap.

Perhaps you've heard of him?

The topic is Thinking Clearly About Performance.

Here's the abstract:


Here's Mr. Millsap's Bio (note the inclusion of the word "teacher", my favorite description):


RSVP here.