Sunday, December 29, 2013

How-to: Upgrade Usage Tracking OBIEE 10g to OBIEE 11.1.1.x

Oracle has provided plenty of documentation on how to upgrade your OBIEE 10g environment to their new 11g platform. They've even outlined the process of configuring a new instance of Usage Tracking on a new 11g environment - outlined here. What I haven't been able to find is the ability to take an existing Usage Tracking environment in OBIEE 10g, and port it over to the new 11g environment. The upgrade assistance tool won't work because the data model for Usage Tracking has actually changed between the two versions.  This guide is going to cover the process of:

  • Taking an existing OBIEE 10g Usage Tracking environment 
  • Making the necessary changes to the Usage Tracking data model
  • Applying the configuration changes in weblogic to enable Usage Tracking in OBIEE 11g

Step 1: Make changes to existing OBIEE 10g Usage Tracking RPD

The data types and nullable check box have changed for nearly every column in Usage Tracking's primary table: S_NQ_ACCT. Using the Admin Tool, modify each column as outlined below:

Step 2: Modify Physical Data Model

I've noticed on some engagements that the NQ_LOGIN_GROUP table may not always be deployed. I don't think this is an 'out of the box' table but rather something manually deployed to track the applicable group of an individual user. Note that as you move to the application role based security model you'll no longer need this table. In the event that you do need this backwards compatibility, you'll need to fix the broken link - which is made evident by the red line that appears when you attempt to display the physical data model:

You'll need to make the physical join on NQ_LOGIN_GROUP.LOGIN = S_NQ_ACCT.USER_NAME

Step 3: Enable Usage Tracking in Enterprise Manager

It's at this step where Usage Tracking implementation differs dramatically between OBIEE 10g and OBIEE 11g. In the 10g platform, Usage Tracking was enabled by modifying the nqsconfig and instanceconfig files, but with the advent of Oracle's Enterprise Manager you must now go through EM's MBeans configuration to enable the tool.  To Oracle's credit, they did provide a detailed guide on how to enable Usage Tracking in EM . Rather than re-invent the wheel, i'm going to defer everyone to the section titled 'Configuring Usage Tracking in Enterprise Manager'. We're past the 'OBIEE 10g' upgrade so Oracle's guide is more than appropriate.

Step 4: Deploy Usage Tracking Tables to Your 11g Database

Most implementations use the DEV_BIPLATFORM that come pre-populated with the S_ETL_DAY, S_ETL_TIME_DAY and S_NQ_ACCT tables, but if your version did not, or you're not using the default schema, run the following scripts located at the following directory:
  1. Oracle_create_nQ_Calendar.sql
  2. Oracle_create_nQ_Clock.sql
  3. Oracle_nQ_Calendar.sql
  4. Oracle_nQ_Clock.sql

Step 5: Confirm Usage Tracking is Enabled

I've created a sample report with the Usage Tracking subject area as outlined below:

When viewing the log generated by the query, you'll see that a query is generated against the Usage Tracking connection pool and no error is thrown!


keywords: obiee 11g, usage tracking,  upgrade usage tracking , enterprise manager

Monday, September 9, 2013

How-to: Automated Web Catalog Deployment in OBIEE 11g

In a previous article I covered the process of how to automatically migrate the repository from a developer environment to production. Equally important is the process of migrating the web catalog.  Most BI Architects are familiar with the ability to archive and unarchive, but there are many unknowns that the Oracle documentation does not cover.

  • What about version control?
  • How do you migrate only select files?
  • How do you migrate object level security?
  • What if there are multiple people editing the web catalog?

A typical web catalog migration path consists of:

  1. Developer checks out or download a local copy of the web catalog from version control
    1. This is only done on an as needed basis
  2. Developer makes changes to specific web catalog files
  3. Developer archives changes
  4. Developer checks out a folder (in this case called from version control that contains all production changes for the specific release
  5. Developer adds their changes to the
  6. Developer updates 'change list' within the to contain path of changed document
  7. Developer checks-in to version control
  8. Deployment script automatically propagates through out each environment as outlined below:

This process uses the '' file as the directory to propagate changes throughout all environments. This is a preferred method because the file only contains changes to the web catalog, and not the entire web catalog itself.

What's in the

The will contain 
  1. Only the modified web catalog objects for your specific release. 
  2. A csv file that contains the destination path of each web catalog object
    1. This is used as input for the shell script we're going to create for automatic deployment
    2. I call this file 'Catalog_Deployment.csv' but it can be renamed if needed
Your file might contain the following:

In the above example I have 5 files:
  1. Financial_Reports.catalog is an archive of the 'Financial Reports' folder which contains multiple reports
  2. Financial_Reports_Dashboard.catalog is an archive of the Financial Reports Dashboard that displays all of the financial reports
  3. HR_Reports.catalog is an archive of the 'HR Reports' folder which contains multiple reports
  4. HR_Reports_Dashboard.catalog is an archive of the HR Reports Dashboard that displays all of the HR reports
  5. Catalog_Deployment.csv is a csv file that contains the target directory path when we unarchive each catalog file (via a shell script)

In the above example, the Catalog_Deployment.csv file would contain the following:

The Catalog_Deployment.csv contains only 2 columns:

  1. The name of the archive file(s) in your (Column A)
  2. The unarchive target directory (Column B)

How do you identify the unarchive target directory (Column B)?

The unarchive target directory (column b) can be found in the 'location' section of the' properties' tab for the specific folder you're trying to archive:

Let's cover the steps required in order for the web catalog to make it out of the developer's box and into assembly test

Step 1.  Check out the file from your version control software

'Check out' is the correct terminology because this will ensure the file is locked and no one can make any changes except for the specific developer. You should now have a locked version of that you can modify.

Step 2. Archive the modified web catalog files

This is a straight forward step that can be achieved through Answers. The archive button can be found in the 'tasks' section as noted below:

Step 3. Modify the file & Re upload to Version Control

Your modifications should include:
  1. The addition of the modified .catalog files
  2. Revisions to the Catalog_Deployment.csv 
Once the changes are made, you can upload back to Version Control and 'check the' back in so other developers can add their changes. The key here is only one developer modifies the file at a time!

Step 4. Pragmatically deploy the web catalog via a deployment shell script

The script we're going to use is going read each row in the Deployment_Catalog.csv file and and use Catalog Manager's script to perform command line based unarchiving.  Oracle has very little documentation on, but think of it as a way to launch Catalog Manager. You can launch Catalog Manager in either GUI mode or command line mode. Using the ' -cmd unarchive' parameters, we're telling Catalog Manager to unarchive in command line mode.

# This file will read the archive files that are unzipped in /tmp/webcatmigration

/bin/dos2unix /tmp/webcatmigration/archive/Catalog_Deployment.csv /tmp/webcatmigration/archive/Catalog_Deployment.csv1
rm /tmp/webcatmigration/archive/Catalog_Deployment.csv
mv /tmp/webcatmigration/archive/Catalog_Deployment.csv1 /tmp/webcatmigration/archive/Catalog_Deployment.csv
while IFS=, read file path
echo ""
echo First Column in Catalog_Deployment - $file
echo Second Column in Catalog_Deployment - $path
echo ""
/export/obiee/11g/instances/instance1/bifoundation/OracleBIPresentationServicesComponent/coreapplication_obips1/catalogmanager/ -cmd unarchive -offline /export/obishare/catalog/webcatalog -inputFile /tmp/webcatmigration/archive/$file -folder "$path"
done < /tmp/webcatmigration/archive/Catalog_Deployment.csv

This script does the following:

  • Converts the .csv file from binary to unix via dos2unix
  • Reads each row in Catalog_Deployment.csv and passes A(N) and B(N) for each row to the $file and $path parameters
A good option for automatic deployment would be to have this script run every Friday after hours as part of the deployment process. You would have one script in each environment for deployment. Ideally the script would be able to read directly from the file server of your version control software.

In summary, we've accomplished the following:
  • Implemented a mechanism to manage web catalog modifications
  • Automated the web catalog deployment process
  • Minimized human error


How often should I back up the entire web catalog?

Notice that the only component of the web catalog being uploaded to version control are the actual changes to the web catalog objects. It is a good idea to take a back up of the entire web catalog prior to the 'go live' of deployment of each 'new release'

What about object level security?

The and archive/unarchive functionality will automatically migrate object level security of each object, but if the object level security is using new application roles that do not exist in the assembly/system/pre-production/production environments - the application roles must first be created in each environment's Enterprise Manager.

Don't forget about the GUIDs!

If you are migrating to a production environment that has different authentication providers than the non-production environment(s) - you must refresh the GUIDs first. You can read how to refresh the GUIDs in this Oracle note - How to refresh GUIDs for OBIEE 11g ? (Doc ID 1564006.1)

keywords: OBIEE 11g, deployment, archive, unarchive, migration, answers, web catalog, webcat

Wednesday, August 14, 2013

How-to: Bridge Tables and Many to Many Relationships Demystified in OBIEE 11g

Bridge tables - entire books have been devoted to this concept, countless blogs write about it, and organizations offer entire classes dedicated to demystifying this idea. Ralph Kimball, creator of Kimball Dimensional Modeling and founder of the Kimball Group has written quite a few great articles discussing the theory of bridge tables.

Yet when researching for comprehensive guides on how to actually implement a bridge table in OBIEE 11g, the documentation available is either:
  • Out of date
    • Contains implementation steps for OBIEE 10g which has since been deprecated
    • Does not contain adequate detail 
      • e.g. missing key steps
This guide is going to outline the basic use case of a many to many relationship, how OBIEE 11g resolves this dilemma and how to successfully implement a bridge table model within the 11g platform.

First thing's first - what is a bridge table and why do we need it?

At its core, bridge table solve the many to many relationship we encounter in many datasets. Many to many relationships in itself are not "bad", but when attempting to conform a data set to a star schema - many to many relationships just do not work. Star schemas assume a one to many (1:N) cardinality from the dimension to the fact. This means "one attribute of a dimension row can be found in many rows of the fact table".

For Example:
  • One job (job dimension) can be performed by many people
    • You would see the same JOB_WID repeating in the fact table
  • One employee (employee dimension) can have many jobs
    • You would see the same EMPLOYEE_WID  repeating in the fact table
  • One call at a call center(ticket dimension) can have many ticket types
    • You would see the same CALL_WID repeating in the fact table
  • One patient (patient dimension) can have many diagnosis
    • You would see the same PATIENT_WID repeating in the fact table

This 1:N cardinality is represented in OBIEE as (using call center/employee example) :
"Cardinality of '1' applied to the dimension and cardinality of 'N' applied to the fact'

But what happens when in the above examples, the cardinality is actually N:N? 

For Example:
  • Many employees can have multiple jobs and each job can be performed by multiple employees
  • Many patients can have multiple diagnosis and each diagnosis can be 'assigned' to many patients
  • Many calls can have multiple call ticket types and each ticket type can belong to multiple calls
This many to many relationship is initially (and incorrectly) represented in OBIEE 11g as:
'Cardinality of '1' is applied to the two dimension tables and cardinality of 'N' is applied to the fact'

Any BI Architect should recognize the above model - it's a traditional star schema! If you stop here and decided to ignore the issue with your dataset and 'hope' OBIEE aggregates the model correctly, you're about to be disappointed.

Why star schemas dont work for N:N cardinality

Consider the following scenario: You're a call center manager and you want to capture the number of calls with positive feedback per employee. You also want to capture the type of calls an employee answers in any given day.

Upon analysis of the requirements you conclude that "each call received by an employee can have many call types and each call type can be answered by multiple employees".

For example:
  • I answer a take a call that is classified as a 'new call', 'urgent', and 'out of state transfer' (three different call types) - this is the "each call received by an employee can have many call types".
  • A colleague also received a phone call that is classified as 'out of state transfer' - this is the 'each call type can be answered by multiple employees"

Now let's put this data in a traditional star schema fact table as modeled below:

1 1 1 300
2 1 2 300
3 1 3 300
4 2 2 500
5 2 3 500
6 3 1 200

You can see in the above data set that:

    • Has 3 different call types
    • Has 300 positive reviews (NUMBER_OF_GOOD_CALLS) 
      • This metric is at the EMPLOYEE level and not the call type level!
    • Has 2 different call types
    • Has 500 positive reviews (NUMBER_OF_GOOD_CALLS)
      • This metric is at the EMPLOYEE level and not the call type level
    • Has 1 different call type
      • Has 200 positive reviews (NUMBER_OF_GOOD_CALLS)
Now you receive a requirement to create a KPI that displays the Number of Good Calls as a stand alone widget.

PROBLEM 1 - Aggregation :
The number of good calls you received based on the above fact table is not 2100 - it's 300 + 500 + 200 = 1000

  • Employee 1 received 300 good calls
  • Employee 2 received 500 good calls
  • Employee 3 received 200 good calls
but due to the many to many cardinality of the data, the star schema duplicates the measures because each employee can take calls of many call types and each call type can be assigned to many employees!

PROBLEM 2 - Grand Totaling:

What if you don't care about aggregates? What if you just want a report that contains the employee, call type and a summation/grand total?

Notice how NUMBER_OF_GOOD_CALLS is repeated across multiple call types and the grand total is still incorrect. It's being duplicated due to the many to many relationship that exists between call type and employee. Furthermore, it paints an incorrect picture that NUMBER_OF_GOOD_CALLS is some how related to CALL_TYPE

How do we resolve this many to many cardinality with a bridge table?

When all is said and done, the incorrectly built star schema:

should be modified to:

Let's break this down:

The bridge table:

This the purpose of the bridge table is to resolve the many to many relationship between the call type and employee. It will contain, at a minimum, the following four columns:
  1. The primary key of the table
  4. The weight factor
The weight factor is what's going to resolve the issue of double counting. 
  • If an employee has 3 call types, there would be 3 rows and the weight factor of each row would be .33
  • If an employee has 10 call types, there would be 10 rows and the weight factor of each row would be .1
In our bridge table data set, we're going to use the same 3 EMPLOYEE_WIDs and create the following:

11 1 1 0.33
12 2 1 0.33
13 3 1 0.33
23 2 2 0.5
24 3 2 0.5
31 1 3 1
You can see from this example that we've taken the N:N dataset in the fact table and moved it into this bridge.

The dimension that is joined to both the fact and bridge

This is a generic dimension that contains the unique EMPLOYEE IDs in your organization's dataset.
For example:
1 1
2 2
3 3
4 4
5 5
6 6
7 7
8 8
9 9
10 10

The dimension that is joined to only the bridge table

This dimension contains all of the possible call types. Note how this table is not physically joined to the fact. This is because this specific dimension (CALL_TYPE) is what's causing the N:N cardinality
For example:
1 Call Type 1
2 Call Type 2
3 Call Type 3
4 Call Type 4
5 Call Type 5
6 Call Type 6
7 Call Type 7
8 Call Type 8
9 Call Type 9
10 Call Type 10

The Fact Table

We've moved the N:N cardinality from the original fact table to the bridge table so the new fact table now contains exactly one row per employee and does not have the CALL_TYPE_WID.

1 1 300
2 2 500
3 3 200

How do we implement this model in OBIEE 11g?

Step 1: Import Tables into Physical Layer

This is always the first step performed when creating a model regardless of its type. In the above example i'm importing four tables:
Step 2: Create the Physical Data Model
Based on our data set above the join conditions would be implemented as follows:

  • 1:N relationship from employee dimension to fact table
  • 1:N relationship from employee dimension to bridge
  • 1:N relationship from call type dimension to bridge
Notice how employee_demo_d is the only dimension that is joined to the fact. w_call_type_d is not joined to the fact because that is the dimension that is causing the many to many relationship issue.

Step 3:  Create the Logical Data Model
The creation of the BMM is where we deviate from our standard build steps of a traditional star schema:

  1. All associated dimension tables referencing the bridge table will be stored in a single BMM table
  2. The single BMM table will have two logical table source

Step 3.1 : Drag the fact table and dimension table that is connected to the fact table into the BMM. 
In our example, we are dragging w_calls_f and w_employee_demo_d into the BMM:

Step 3.2: Create a 2nd LTS in the existing dimension table

  1. Right click W_EMPLOYEE_DEMO_D -> New Object -> New Logical Table Source
  2. Name it 'Bridge'
  3. Add W_BRIDGE_D and W_CALLTYPE_DEMO_D (the two dimensions not directly joined to the fact table) under the 'Map to these tables' section

  1. Next add the remaining dimension columns from W_CALLTYPE_DEMO_D and W_BRIDGE_DEMO_D to the Dimension table in the BMM

Step 3.3: Create a level-based dimension hierarchy for the dimension BMM
  1. This step should be completed whether or not the schema is a star or bridge

Step 3.4: Confirm the BMM model has a 1:N relationship from the dimension to fact
Step 3.5: Set aggregation rule of NUMBER_OF_GOOD_CALLS to sum 
All measures in the BMM must have a mathematical operation applied to the column

Step 3.5: Set the Content level of the dimension table to 'detail' in within the LTS of the fact table
Again, this is something that should always take place regardless of the type of model

Step 4: Create the Presentation Layer
This part is straight forward, just drag the folders from the BMM into the new subject area:

The moment of truth
So why did we go through this elaborate exercise again? To fix the aggregation issues we were having with NUMBER_OF_GOOD_CALLS due to the N:N cardinality of the data set. Let's create that 'standalone KPI' Number of Good Calls:

Notice how the metric correctly sums to 1000. Let's check the back end physical query to confirm:
Notice how it's hitting the fact table and not the bridge or the call type dimension. 
But what about the weight factor?
Let's go back to the scenario where we want to compare across dimensions joined via the bridge table (EMPLOYEE and CALL_TYPE):
  • When creating a report that uses a measure from the fact table, a dimension value from the the employee table, and a dimension value from the table that causes the N:N cardinality - you need to use the weight factor to make sure your measure isn't getting double or triple counted:

  • Notice column is using the the NUMBER_OF_GOOD_CALLS multiplied by the WEIGHT factor in column 2
  • Each row in column 1 correctly represents the NUMBER_OF_GOOD_CALLS in the fact table despite having the repeated values of multiple call types
  • Note the aggregation of grand total sums to 997. This is because the weight factor is rounded to the 2nd decimal for EMPLOYEE_WID = 1 (.33%)
In order for grand totaling to work correctly with bridge table measures that use weight facts you must set the aggregation rule of the column (in this case column 1) to sum within Answers:

So what did we accomplish in this guide?
  • A basic understanding of many to many (N:N) cardinality
  • A basic understanding of why the star schema won't work for N:N cardinality
  • How to resolve the cardinality issue with a bridge table
  • How to implement a bridge table in OBIEE 11g

keywords: bridge table, cardinality, many-to-many, OBIEE 11g, helper table, answers, analytics, aggregation, LTS, measures, kimball

Monday, August 12, 2013

FYI: GoURL Primer with OBIEE 11g

The Oracle BI Presentation Services Go URL command is for use in incorporating specific Oracle Business Intelligence results into external portals or applications. The Go URL is used when you add a result to your favorites, or add a link to a request to your dashboard or an external Web site. It has a number of forms and optional arguments that can be used to control its behavior.

Oracle has provided limited documentation on goURL parameters and with use cases that require end users to navigate to specific dashboards or request dashboards be embedded into pre-existing web pages, i've decided to create a Primer on commonly used GoURL functionality:

Common goURL Parameters:

NQUser&NQUser=xUser ID
Path&Path=xPath of the answer to execute. You can find it in the properties page (Answers/Manage Catalog/ and click on the properties icon (a little hand) of the object that you want. See picture above
Link Options&Options=xThe x can be one or more of the following letters:
* m : Modify Request
* f : Printer Friendly
* d : Download to Excel
* r : Refresh Results
Printer Friendly&Action=printResults are in a printer-friendly format, without the paging controls, hot links, and so on.
Passing Filters&Action=NavigateTo apply filters to the answer (see section below on Passing Filters)
Application Friendly&Action=Extract
Results are displayed in an application-friendly format, such as for Microsoft Excel, without the paging control, hot links, and so on. The Extract action also acts as a Navigate action (read Passing Filters to the Oracle BI Presentation Services Go URL Through a URL (Navigation)) so you can filter the results that are returned by the call.
Specific View&ViewName=xThis shows an individual result view rather than the default compound view
Specific View&ViewID=go~xThis shows an individual result view rather than the default compound view where x is the name of the view
Specific Style&Style=xThis shows the results using a specified style. If the style does not exist, the default is used.
Result Format&Format=xThis controls the format of the results. This is the format, where x can be xml, html, txt (tab separator), csv (comma separator)
File Extension&Extension=.csvThis controls the file extension of the download file
Language&Lang=frThis controls the language of the report. The value permitted are the values of weblanguage
done&done=portalPagesparameter allows you to create a return link equal to a path that you have specified
Passing Filters

You can use the &Action=Navigate in conjuction with the with the parameters below to pass filters directly to a report via goURL. Use &P0=n where n equals the number of parameters you wish to filter and P6 with the one or more of the operators below:

eqEqual to or in.
neqNot equal to or not in.
ltLess than.
gtGreater than.
geGreater than or equal to.
leLess than or equal to.
bwithBegins with.
ewithEnds with.
canyContains any (of the values in &P3).
callContains all (of the values in &P3).
likeYou need to type %25 in place of the usual % wildcard. See the examples that follow.
top&P3 contains 1+n, where n is the number of top items to display.
bottom&P3 contains 1+n, where n is the number of bottom items to display.
betBetween (&P3 must have two values).
nullIs null (&P3 must be 0 or omitted).
nnulIs not null (&P3 must be 0 or omitted).
&P2=ttt.cccIn this parameter, ttt is the table name and ccc is the column name.
If the table or column contains spaces, it must be quoted with double-quotes.
 Spaces should be escaped as %20, for example, Measures."Dollar%20Sales".
&P3=n+xxx+yyy+...+zzzIn this parameter, n is the number of values, and xxx, yyy, and zzz are the actual values.
Note: If the value of P3 begins with a numeric character,
the entire value must be enclosed in quotes.
example: saw.dll?Go&Path=/Shared/Test/SB2&Action=Navigate&P0=1&P1=top

Passing Filters Examples:

his returns records for the East and Central regions:
This returns records for like Regions E....t:
This returns the top two regions by dollars sold:
This is an example where the number of arguments is not included in the syntax:
This returns records with between 2,000,000 and 2,500,000 in sales:
This returns records for Regions beginning with the letter E:
This returns records for Regions containing the letter E and having more than 20 million in sales:
Generating SQL Statements using goURL:

The Go URL command can be used to issue Oracle Business Intelligence SQL. These forms of the Go URL return tabular results. The basic options from &Style= and &Options= can be used here as well.
To issue Oracle Business Intelligence's simplified SQL, include the escaped SQL as a parameter to the Go URL. For example:
where the FROM clause is the name of the Subject Area to query.

Alternatively, the command IssueRawSQL can be used to bypass the Web processing and issue SQL directly against the BI Server.
keywords: OBIEE 11g, goURL, answers, ad-hoc, analysis, analytics

How-to: Automatically Refreshing Dashboards in OBIEE 11g via iFrame Bursting- The Flat Screen TV Use Case

With production costs associated to flat screen LCD/plasma/LED TVs dropping into the 100s of dollars over the past few years, these TVs are no longer relegated to the home theater space. Now enterprises can utilize these products as part of their campaign to bring analytics to the forefront of their business. Consider the following scenarios:
  • A call center needs a mechanism to communicate performance metrics to the office without daily emails, meetings and paper-printed dashboards
  • An IT organization needs to track hardware utilization 'in real time' with passive notifications activated during peaks and irregular activities
  • The firm's testing team needs the ability to communicate defect and resolution metrics against service level agreements across multiple teams
Short of daily emails and unnecessary meetings, it was up to the individual worker to take the initiative to stay informed. The above scenarios can be effectively resolved through the use of 'Dashboard Displays' on flat screen TVs strategically placed throughout the office to keep leadership informed and team members accountable.

When we think of how OBIEE is used, the typical scenario that comes to mind is a user logging into the analytics front page. While little documentation exists, it is certainly possible to transform OBIEE from a 'user-based interactive analytics tool' to a 'dashboard style display fitted to large monitors'.

The main issue OBIEE Architects face is how resolve the issue of stale data across the technology stack. Stale data can exist within the BI Server, BI Presentation Services and even in the browser. Each component has its own resolution technique as well as potential issues. When a dashboard is displayed on a static monitor, a mechanism must be in place to ensure it's refreshed and timed correctly with existing agents.

 This guide is going to outline the steps required to implement OBIEE for the above scenarios. The technology used includes:
  1. iFrame Bursting (disabled by default in OBIEE 11g)
  2. Oracle WebLogic to redeploy modified XML components
  3. goURL
  4. HTML  (Meta Tags)
  5. Oracle BI
Some blogs on the web recommend using javascript to achieve this but I do not recommend this approach as javascript execution is based on the permission of the browser and has the potential to be disabled. At the very least a pop up box will appear asking the user if they authorize the javascript function to execute - not a pleasant end user experience.

Furthermore, Oracle has acknowledged that the mechanism they developed for auto-refresh does not work and no known resolution exists - see Doc ID 1416002.1 and  Bug 13058029. What is interesting is that Oracle's method used to work in OBIEE 10g!

Step 1: Enable iFrame Bursting

By default, OBIEE 11g server does not allow dashboards to be displayed in an iFrame. The reason this is done is to better secure the dashboards and prevent hackers from using an iFrame to overlay malicious content.

Step 1.1. Navigate to export/obiee/11g/instances/instance1/config/OracleBIPresentationServicesComponent/coreapplication_obips1 and modify instanceconfig.xml 

Add the following red code to the file:
Step 1.2 Download analytics.ear

Navigate to:
/export/obiee/11g/Oracle_BI1/bifoundation/jee and download analytics.ear
The analytics.ear file contains two war files that house the web.xml that needs to be modified.

Step 1.3 Modify web.xml

Decompress the analytics.war and analytics-ws.war file using 7zip and modify the web.xml in each war file by adding the red code to the file:
<url-pattern>/RelatedContent</url-pattern> </servlet-mapping>
<context-param> <param-name></param-name> <param-value>never</param-value> </context-param><login-config> <auth-method>CLIENT-CERT</auth-method> </login-config>
Step 1.4 Upload & Recompress the war files
Upload the two war files back to /export/obiee/11g/Oracle_BI1/bifoundation/jee

Step 1.4.1 Recompress analytics.war

Navigate to:
 jar cvf0 ../analytics.war 
(Note: Include the space and period at the end of the command)
Navigate up one level to:
Delete the analytics folder so that the only files that remain in that directory are: 
(1) META_INF, (2) analytics.war, and (3) analytics-ws

Step 1.4.2 Recompress analytics-ws.war 

Navigate to:
jar cvf0 ../analytics-ws.war . (Note: Include the space and period at the end of the command)
Navigate up one level to:
Delete the analytics-ws folder so that the only files that remain in that directory are:
 (1) META_INF, (2) analytics.war, and (3) analytics-ws.war

Step 1.5 Recompress the analytics.ear file

Navigate to:

 Rename the old analytics.ear file as a backup

Execute: jar –cf analytics.ear –C analytics . (Note: Include the space and period at the end of the command)e)Delete the analytics folder.

Make sure that the analytics.ear file was produced in the export/obiee/11g/Oracle_BI1/bifoundation/jee/ folder. 
Delete the analytics folder.

Step 2: Restart BI Services

This can be achieved by either running opmnctl stopall via command line or through Enterprise Manager as depicted below:

Step 3: Create a test HTML file with the embedded iFrame:

Open a text editor of your choice and paste in the following:

<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Frameset//EN" "">
<TITLE>Important Dashboard</TITLE>
<META http-equiv=no-cache>
<META content=900 http-equiv=refresh>

<IFRAME style="Z-INDEX: 998; POSITION: absolute; WIDTH: 100%; HEIGHT: 100%; TOP: 0px; LEFT: 0px" src="" frameBorder=no scrolling=auto> </IFRAME>

Note the two important components highlighted in red:

  1. The Meta tags forcing the automatic refresh every 900 seconds. 
  2. The iFrame that embeds the URL of your dashboard
So at the end of the day what do we have?

  1. The ability to display dashboards on a static device (large monitor) that will have little human interaction but requires up to second (or in this case every 900 seconds) data
  2. The ability to embed a dashboard in an iFrame.

keywords:  OBIEE 11g, iFrame, auto-refresh, answers, web.xml, instanceconfig.xml, goURL, analytics

Monday, July 29, 2013

How-to: Image Referencing with OBIEE 11g

In a typical OBIEE engagement, the client may want to utilize out of the box or custom images within various dashboards and reports. This requirement leads to many open questions, including:

  1. Where are these images located?
  2. How do I embed the image into a dashboard or report?
  3. How do I maintain the integrity of the image URL across multiple environments?
Let's break down each question one by one:

Where are these images located?

All images are stored in the 'browser look and feel plus' folder of the BI Server, you've probably seen this notated as 's_bflap'. This folder exists in two locations and it is critical that any image you upload be housed in both:
  • Oracle_BI1\bifoundation\web\app\res\s_blafp\images
  • user_projects\domains\bifoundation_domain\servers\bi_server1\tmp\_WL_user\analytics_11.1.1\7dezjl\war\res\s_blafp\images

How do I embed the image into a dashboard or report?

OBIEE 11g has a little known feature called 'fmap' which can be used to display an image based on the relative URL of the image. Little documentation exists on it other than a few notes released by Oracle which include:
  • How To Display Custom Images Using Fmap In OBIEE 11g (Doc ID 1352485.1)
  • Image FMAP on Linux (Doc ID 491154.1)

How do I maintain the intregrity of the image URL across multiple environments?

Here is where things get tricky due to the lack of documentation that exists.  Let's say you want to use the image 'report_good_percentage.jpg' located in the s_blafp folder:

So as outlined in Oracle's documentation you use 'fmap:report_good_percentage.jpg' or even 'fmap:images/report_good_percentage.jpg', but to your dismay all you see is a broken image link:


It is important to remember that fmap displays the image of the relative URL. So what does relative mean? What is 'it' relative to? In regards to fmap, the relative URL is the root directory of the analytics web server, which in OBIEE 11g is:

Which makes sense if you understand how applications are deployed in weblogic. The presence of the WEB-INF directory in the aforementioned folder is how Weblogic determines if a folder if a deployable application directory.

So - if we work under the assumption that the above folder is indeed the root directory, then it we now know why the image returns a broken link, report_good_percentange.jpg is not stored in the 'root' directory of the analytics web server, it is actually stored in:
Let's update the fmap relative url to correctly reference report_good_percentage.jpg by modifying it to:

Why does fmap STILL not work?

Let's take a look at the URL that's actually being generated:

Notice anything funny? Why is OBIEE 11g adding a 'Missing_' folder to the URL directory? Countless bloggers have theorized this as a bug in OBIEE and some even suggest making a 'Missing_' folder in the root directory of the analytics web server. I don't think that is the best approach because as you deploy this application across multiple servers, you'll have to make sure all environments have that same folder. Keep it simple right?

We can resolve this by modifying the fmap url to revert one directory closer to its root by using '/..':
Let's check the URL being generated just to make sure:

The image displays, and the 'Missing_' folder is no where to be found.  If your requirements have extensive image customizations, perhaps changing the entire look and feel, I recommend deploying an entirely new skin as outlined in Oracle Note: How to Use Custom Images in OBIEE (Doc ID 1484623.1)

keywords: fmap, answers, OBIEE 11g, image customization