Friday Apr 17, 2015

Having Trouble Publishing Projects?

Because the P6 Extended Schema is the lifeline of data for Primavera Analytics it is an important topic to cover.
There are two white papers which go into depths about the feature, how to setup, how it works, and some advanced configurations and techniques.

The key to having project publish is a job service called the Arbiter. If you query the jobsvc table as admuser you will see this job.

select * from jobsvc where job_type = 'JT_ProjectArbiter';

When this job runs it looks through your projects to see which ones have changes, which have crossed the time threshold, and which have been marked as 'Publish Now' and
adds these projects to the queue.  Without this service running no projects would be published.

In a scenario where your database has crashed or other environmental issues may cause this job to be stuck in a running or bad state you will want to get it running again as quickly as
possible. One way would be to put a monitor on this row and trigger a database alert if it is stuck in running state for an extended period of time or does not run on the designated schedule.

If the arbiter is running ok but some projects might not be getting processed there are a few reasons:

1- There are no changes on the project so there is no need to process it.  As discussed in an earlier blog not all changes on the project count as a change, changing a code or udf assignment wouldn't count as a change.

2- There is a setting called 'Publish Idle Projects.'  This really only occurs when you first turn on publication.  At that time it will go through all projects with enable publication
checked and process them, when it is completed this is turned off.  It is turned off by altering a value - ProjectNonDemandStopKey - in the Settings table to zero. If this setting
is zero then new projects will not be published until they have changes, have been scheduled, or have been opted in by choosing 'Publish Now.'  You can turn this back on from the database
by running the following as admuser:

update SETTINGS set setting_value = (select max(proj_id) from project)
where SETTING_NAME = 'ProjectNonDemandStopKey';

This will force the arbiter to go through and add any unpublished projects into the queue.  Once it has completed those projects it will set itself back to zero and turn off.

3- Is a rare scenario where none of the above conditions have been met and you just want to force in all unpublished projects.  You could run the following:

update project set px_next_date = sysdate where proj_id in
(select proj_id from project
and PX_LAST_UPDATE_DATE is null;

We will have more blogs discussing the Project and Global publication services. This one is intended to give a little more background on the services.

Monday Mar 23, 2015

When Configuring Multiple Data Sources...

When you are setting up additional data sources to be applied to your STAR schema (For example - multiple P6 databases being combined into a single STAR schema) there are a few things you need to make sure are configured.

1- New installation directory
    For this new data source it should have it's own installation directory with a new folder name (especially if on the same machine).
    Edit the following files and change the path for the new staretl directory:
        config.cmd or sh
        \etl\common\runSubstitution.cmd or sh
        \res\ - change any paths that are incorrect. If there is a filter define it for DS2 (star.project.filter.ds2=filtername).
        staretl.bat or sh
            -add the data source number into this file so each time it executes it runs with the proper data source
                call staretl.bat "-s2" %1....        

    In the scripts folder, locate staretl.bat or .sh. Edit the file, search for 'primary_link', remove the step where you find this from staretl then save.  The reason for this is the script this step runs drops and recreates database link 01 each time.  It also has parameters which could be populated with another source if executed.

2- Create new database link - DSLINK02
    Create a new database link called DSLINK02 for the 2nd data source. The same pattern would repeat for additional data sources.  This new DSLINK should have the connection defined to the PXRPTUSER or filtered view for this new source.

It is critical to make sure these steps are followed when creating a new datasource and running the ETL processes. If steps are missing the ETL's could run against the wrong source for the specific data source. These steps are also documented in the Installation and Configuration guide.

Friday Mar 20, 2015

I Created an ETL Schedule Through the Web Configuration, How Do I turn the Schedule Off?

In P6 Analytics 3.3 a web configuration utility that can be deployed in Weblogic was introduced.  Through this utility you can define an ETL schedule or run the ETL on demand. This web utility will also create a queue to prevent ETL processes from colliding. If you have multiple data sources you can define a schedule for each and the queue will be built to so the ETL process will wait for the previous or current ETL to finish before starting the next.
When creating the schedule you can choose to run daily at a designated time or define which days of the week to run and at what time each day. When you define your schedule it is actually written into your ETL installation directory (staretl\res) in a properties file ( Any changes you make to your schedule will be reflected here.  You can only have one schedule per ETL installation.

What if you created a schedule to run every week on Monday, Wednesday, Friday at 10pm and now you want to run just daily?
All you need to do is make a change on the Schedule page in the web application. By unchecking weekly and selecting daily, selecting your time and clicking 'Schedule ETL' you will have overwritten the previous schedule.  You can verify this by checking the

What if you want to turn off all scheduled ETL's?
You can choose Weekly and leave all days unchecked.  This will write to the schedule.propertie to turn off for all options, daily and each day.

Monday Jan 26, 2015

Summary Data and the Summarizer, What is Needed for P6 Analytics?

In P6 there is the Summarizer, which runs scheduled jobs to do calculations and aggregations of data. This data is visible when looking at Projects and EPS in the P6 client or web applications.
There are also the Global Services associated with the Extended Schema. In the Global Services there are jobs for Enterprise Summaries, Enterprise Data, Resource data, and Security. These jobs are completely separate and do not depend on or use any data from the Summarizer. The Summarizer and P6 Extended Schema services are two completely separate entities. P6 Analytics only uses the data coming from the Extended Schema, therefore it is not necessary to run the Summarizer to populate data for P6 Analytics. Only the Global services and Project Publication services need to be run.

Friday Jan 09, 2015

Capturing Historical Financial Periods for Analytics

In P6 Analytics you can use Financial Periods as an element of time similar to Week, Month, Year, etc.  Financial Periods are defined in P6.  They can be a week, two weeks, a month, it is a user defined period.
The financial periods are available where the time dimension is available but can react differently dependant on the subject area that is being used. In this blog we will cover two main areas where this might be used and how the two subject areas will work differently.

Primavera - Activity

This subject area is not historical.  This data is as of the last ETL run. The data in this subject area comes from the Activity Spreads.  If you have spread data that falls into these financial period buckets then it will be shown.  If you do not then certain financial periods will not be shown for this project. Remember once a fact is added into the Analysis, or you join to another dimension hence joining to the fact table, data will be filtered out to honor the data in the fact tables.  Analytics is fact driven. You must have a fact to fit the criteria for it to be displayed.

Primavera - Project History

This subject is historical.  As we discussed with the Primavera - Activity subject area, this is still fact driven.  The change in behavior is that these facts are all about being historically captured, this is NOT using spread data it is only using historical data. For example, you have a project that ranges from 1/1/2013 to 1/1/2016.  You have financial periods established for every two weeks during this time period.  Today is Jan 9th, 2015.  If you are just setting up P6 Analytics today and have never captured any history up to this point when you run the ETL process you will only see data for the financial period of Jan 1 - Jan 15th 2015.  The historical facts are added into the specified buckets of the day the ETL is run and what bucket that falls into.  If you have history for Week, Month, Quarter, and a Financial Period (every two weeks) you would see in the w_project_history_f a period_start and period_end_date like the image below:

Because you only have facts for those time periods, you would only see those time periods were historical data was captured. If you only have history captured starting today (Jan 9th) you would only see data that falls into that bucket.  As time goes and you continue to capture history you will see additional buckets of history be stored and made available for viewing. 

Thing to remember

Here are a few items to remember to do in P6 to make sure your projects are setup to capture history for financial periods. In the P6 Extended Schema remember if you are defining new financial period buckets to run the Enterprise Data global service so this data is calculated and stored in the Extended Schema
and available to be pulled over during the ETL. Projects must be published to calculate the spread data and changes to the project so they can be pulled over into STAR schema and available in the Primavera - Activity and Primavera - Project History subject areas. History must be enabled for the projects where you want to have history recorded.  This is off by default. When setting the history interval it will need to be set to Financial Periods.

Monday Dec 22, 2014

How Could I Get a Project to Publish After a Certain Amount of Activity Code Changes?

Following up on the previous blog discussing why making a change to a code assignment isn't enough to trigger the publication of a project, here is an option if there is a business need around a certain P6 object that gets updated often and is considered as important as the tables we currently count changes for (PROJWBS (WBS rows), TASK (Activities), TASKRSRC (Resource Assignments), and TASKPRED (Relationships)) which determine if a project needs to be refreshed.  

Say activity codes are changed very often on your projects. Maybe this is a way of moving activities into different phases or an important statusing practice. Because the Project publication infrastructure is based on changes, you would want to queue up a project based on changes as well. In this SQL example below we are going to update the px_next_date on the row for this project in the PROJECT table.  Px_next_date when updated with a date later than the last run of the arbiter (project publication) service, it will cause this project to be added to the queue.

update project set px_next_date = sysdate where proj_id in
--timestamp, and trigger a publication for projects in the following query
(select p.proj_id FROM
(select proj_id,  count(*) COUNT from taskactv
where update_date > (sysdate -1)
--in this example where there has been an update in the last day, this could be extended to 7 days or any other time choice
group by proj_id
order by proj_id) p
where p.COUNT > 50)
---when there is a total of 50+ changes publish this project

In this example we are saying at the time of running this SQL, go get a count (per project) for the amount of activity code updates within the last 1 day. And if that count is greater than 50 mark this project to be published.  You could change the numbers to better fit your needs.  You could do a similar update on other project
related objects, this is just a high level example (this is not meant to be production code).

Thursday Dec 11, 2014

Why Aren't My Code Changes Being Reflected in Analytics?

The scenario is, you logged into P6, you made a series of activity code changes on your project.  An hour or two later you ran the ETL process, expecting to see the codes changes reflected when you viewed them in OBI or directly in your Star schema.  However they are not updated, why?

Let's walk through the process of how the data gets there and why codes are different than updating activities directly.

Step 1:  Project Publication - what counts as a change.

You are dependent on the thresholds being crossed for updating projects in the Extended schema (See earlier blog for more info on thresholds). Basically changes to a project are counted, then if you reach the threshold (say 100 changes) or the time threshold (say 24hrs. and 1 change) the project will be published.  However items like code assignments do not count as a change. The only updated areas that count as a change are updates to the following tables:

PROJWBS (WBS rows), TASK (Activities), TASKRSRC (Resource Assignments), and TASKPRED (Relationships)

Step 2: Publishing the Project

The design behind this is these are the most commonly updated areas in a schedule.  To throttle the system and avoid overloading the queue with unnecessary updates these are the core areas that drive updates.  If you are making major changes to codes it is good to be aware of this and you can choose to 'Publish Now' which will then force a publication of your project and capture your code changes.

Step 3: Run the ETL

OK now that we have the Extended Schema (Project publication) side understood and worked out you can run the ETL as normal.  Remember the Extended Schema is the heart of all the data that makes it into P6 Analytics.  P6 Analytics and your STAR schema is only as up to date as your Extended Schema.  

Friday Nov 21, 2014

History Selections (Part 2)

In the last blog we discussed how the history selections you make can effect the size of your Star schema.  Now let's talk about how we make the selections and what they mean.  In P6 go to Projects, then the EPS view.  Right click and choose Project Preferences, and click on Analytics.

Here you will be presented with History Level and History Interval.   History Level is the most important setting here.

Choosing Activity:

-This will populate daily activity level history for the Primavera-Activity History subject area.  This is always daily.  There is no setting for Activity History to be a granularity other than daily.

-When you choose Activity History the History Interval drop down is still available.  This is for choosing the granularity for WBS and Project History, which are turned on by default when you choose Activity Level History.  

-Slowly Changing Dimensions will be turned on for changes to this Project.

Choosing WBS:

-If you choose WBS for History Interval this will also turn on Project Level History and will make these available in the Primavera -Project History subject area.

-The History Interval selection will now affect both WBS and Project Level History.

-Activity Level history will be off as well as Slowly Changing Dimensions. 

Choosing Project: 

-If you choose Project for History Interval this will turn on only Project Level History and will make it available in the Primavera -Project History subject area.

-The History Interval selection will now Project Level History.

-Activity and WBS Level history will be off as well as Slowly Changing Dimensions. 

Thursday Oct 30, 2014

How Historical Selections Effects the Size of Star Schema?

For P6 Reporting Database, the Planning and Sizing guide covers expectations for the size of your Star schema based on the size of your P6 PMDB and gives recommendations.  History is discussed in this document as well, but one thing to go a little deeper on is how the selection of history per project can have an effect on your ETL process as well as the size of your Star schema.

If you select Activity History on a project this will:
- record any change on the Project, Activity, and any related Dimension each time the ETL is run (ex. change in activity name). This enables Slowly Changing Dimensions for rows related to this project.
- Activity history will be recorded in the w_activity_history_f on a daily level.  A row will be captured for each activity for each day the ETL process is run. Depending on the amount of projects you have opted in this could grow quite quickly.  This is why partitioning is a requirement if you are using Activity level history.  
- Choosing Activity level history also opts you in for WBS and Project level history at the selected granularity (Weekly, Monthly). Recording a row for each WBS and Project each time the ETL process is run and bucketing the results into the chosen granularity.

Because choosing Activity History can have a such a dramatic effect on the size and growth of your Star Schema it is suggested that this be a small subset of your projects. Be aware of your partitioning selections as this will help with performance.

For determining the amount of rows for Activity History, take number of projects opted in times number of activities in each project times the number of days in the duration of the project times number of ETL runs. This will give you a ballpark idea on the amount of rows that will be added per project.  A similar calculation can be done for WBS history as well, adjusting for number of WBS rows and selected granularity. 

In summary, be cautious of turning on Activity History on a project unless you absolutely need daily level history per activity.  If the project is on, once the project has ended turn off the Activity History on the project.  Keeping project controls accurate and up to date can help improve the ETL process and control Star schema. 

Friday Oct 03, 2014

P6 Analytics and P6 Reporting Database Security Enforcement

In P6 Analytics row level security is used to enforce data security. The security is calculated in P6 by the extended schema global security service. This means that P6 Analytics uses the same security defined in P6. If you gave a user access to a project in P6 they would have access to it in P6 Analytics. It is calculated, stored, and used across both products.  The calculated security data for projects, resources, and costs are pulled over during the ETL process. This is stored in the STAR schema and when querying the STAR schema for data using OBI there is a validation of the user in OBI against a user that should exist from P6. When the user is identified the data is filter based on the calculated out data they have access to. This is handled using a package in the STAR schema - SECPAC. Security policies for row level security are applied during the ETL process and can be found in the \scripts folder. This enables enforcement across the Facts and Dimensions. Using OBI and allowing the ETL process to handle the enforcement is the easiest way.

However, if you do not use OBI but still wanted to use the STAR schema for data and use the security enforcement you can still use just the P6 Reporting Database.  You would setup your STAR schema and ETL process just as you would if you had the full P6 Analytics.  Once the ETL is completed you will need to execute the SECPAC and pass a username to set the context. In a session with this defined, this user would only see the data they had access to as defined in P6. This would be an additional step and need to be pre-defined in any process to access the STAR data.  Using OBI is a much easier option and you can take advantage of all the benefits of OBI in creating Analysis and Dashboards but it is still possible to setup STAR and use the calculated P6 security even without OBI in your environment.

Wednesday Oct 01, 2014

Incorporating an HTML Weather Feed into an OBIEE Analysis

We've created a whitepaper that details how to embed an HTML weather feed within an OBIEE analysis, and then integrate that weather feed with a spatial view. The end result is an interactive analysis that displays the weather for the location selected in the spatial map. To download the white paper, follow the link below:

P6 Analytics: Weather Feed in OBIEE

Wednesday Sep 24, 2014

UDF's: what is available and where?

This blog covers what is available for User Defined Fields as of P6 Analytics version 3.3. 

Areas available:

Resource Assignment

These 5 are the Subject areas available for User Defined Fields. Each User Defined Field has a subject area for use with UDF's that are treated as Facts (Cost and Number).

Ex. Primavera - Project User Defined Fields.  

Types of UDFs:

Date and Text - date and text udf's are treated as dimensional values and are grouped into the specific area in each subject area containing that dimension. 
For example, in the Primavera - Activity subject area, under Project you will see UDF-Date and UDF-Text.  These will be available in Analysis in these subject areas. Because Date and Text are dimensional they are also available historically as a slowly changing dimension value when activity history is activated for that project.  Activity History should only be enabled for a small subset of projects because of the impact they will have on database growth.  Please see Planning and Sizing guide for more information.

Number and Cost - number and cost are treated as facts.  They will be the core of each UDF subject areas. For example, Primavera - Project User Defined Fields will have number and cost as the fields available in the Facts section.  The date and text UDF's will be available as dimensional values same as in other subject areas. Because these are facts they are not part of the slowly changing dimensions and do not have historical records. 

Friday Aug 08, 2014

Change to Logging for New Web Configuration Utility

Since P6 Reporting Database 2.0 and Analytics 1.0 the ETL process log has been appending during the ETL runs. In 3.2 staretlprocess.log will contain all ETL runs. We received customer feedback on a desire to have a log for each ETL run.  With the web configuration utility added in 3.3 this was implemented.  In the new configuration tool you can schedule your ETL runs and view the results and logs from the the web config. The logs were separated out so you can see the log for each individual run.  This also gives you the option to clean up old logs from your system directory for a given time period.  

The existing (non web based) configuration utility will still do the logging based on appending each run to the same log if this is the way you choose to have your ETL process logged. However you should choose one method, either the existing configuration utility or the web based configuration utility and not use both.  The new web based config can build up a queue of ETL jobs and avoids any kind of collision. This is a major advantage for those using multiple data sources.  You can queue up multiple runs and they will handle the execution.  The existing configuration utility does not have a queue mechanism for ETL runs, they are controlled by cron jobs or manually being kicked off so collisions there are possible and you can not have more than one ETL process running at a time. Also because of the different logging methods you should not be running ETL's from both methods because the web configuration utility will not know of the logs for ETL runs kicked off from the STAR installation directory.  Each log from the new queue method will have a time stamp and unique name on it for easy identification. 

Increase to Default Values of Codes and UDFs in P6 Analytics

In the P6 Analytics 3.3, the default value of all UDF's (subject area (Project, Activity, etc.) and type (Date, Text, etc.)) were increased from 20 to 40.  This includes the RPD changes.  When choosing your UDF's from the configuration utility these changes will automatically be reflected in the schema and no changes will be needed in the RPD until you hit 40.   We also increased the default value of Codes from 20 to 50.  Same updates apply as with UDFs where all schema and RPD changes are automatically reflected up to 50. If adding greater than 50 for Codes or 40 for UDF's the appropriate RPD changes will need to be made to add additional values but all schema side changes will be handled by the ETL process.

Monday Jul 28, 2014

Deleted data showing in History?

In the Star ETL process there are 2 main types of updates.  There is the current data (dimensions and facts) and there is the historical data (_hd, hf, history) - see

The current data is overwritten every ETL update.  This will keep your data as up to date as the last ETL.  If you worked on a project, deleted a project, updated tasks, deleted tasks, scheduled - that will all be reflected. In the Historical tables, these changes once captured will always remain.  For example, you deleted a project.  In your list of projects this will still be available. Once the data is captured it is always in the database. This way you truly have a historical perspective.   

In P6 Analytics when viewing Historical data you may want to filter out this data.  You can do so by adding a code to these values and set a value of 'Deleted' then apply this filter. If you truly want to see just the most recent data the Primavera - Activity subject area may be the subject area you want to report on.  The historical subject areas is more for trending and seeing changes over time. In a future blog we will discuss how to add a delete_flag to allow for easier filtering in P6 Analytics. The main take away from this entry is once the ETL is run the data is captured as is and always available in the database if it needs to be reviewed. There was a high demand for being able to see all data, even if deleted, to help show the total historical picture of the life of a project. 


Provide new information on Primavera Analytics and Data Warehouse


« February 2016