Skip navigation.

Rittman Mead Consulting

Syndicate content
Delivering Oracle Business Intelligence
Updated: 8 hours 47 min ago

Introducing Oracle Big Data Discovery Part 3: Data Exploration and Visualization

Thu, 2015-02-26 17:08

In the first two posts in this series, we looked at what Oracle Big Data Discovery is and how you can use it to sample, cleanse and then catalog data in your Hadoop-based data reservoir. At the end of that second post we’d loaded some webserver log data into BDD, and then uploaded some additional reference data that we then joined to the log file dataset to provide descriptive attributes to add to the base log activity. Once you’ve loaded the datasets into BDD you can do some basic searching and graphing of your data directly from the “Explore” part o the interface, selecting and locating attribute values from the search bar and displaying individual attributes in the “Scratchpad” area.


With Big Data Discovery though you can go one step further and build complete applications to search and analyse your data, using the “Discover” part of the application. Using this feature you can add one or more charts to a dashboard page that go much further than the simple data visualisations you get on the Explore part of the application, based on the chart types and UI interactions that you first saw in Oracle Endeca Information Discovery Studio.


Components you can add include thematic maps, summary bars (like OBIEE’s performance tiles, but for multiple measures), various bar, line and bubble charts, all of which can then be faceted-searched using an OEID-like search component.


Each visualisation component is tied to a particular “view” that points to one or more underlying BDD datasets – samples of the full dataset held in the Hadoop cluster stored in the Endeca Server-based DGraph engine. For example, the thematic map above was created against the post comments dataset, with the theme colours defined using the number of comments metric and each country defined by a country name attribute derived from the calling host IP address.


Views are auto-generated by BDD when you import a dataset, or when you join two or more datasets together. You can also use the Endeca EQL language to define your own views using a SQL-type language, and then define which columns represent attributes, which ones are metrics (measures) and how those metrics are aggregated.


Like OEID before it, Big Data Discovery isn’t a substitute for a regular BI tool like OBIEE – beyond simple charts and visualizations its tricky to create more complex data selections, drill-paths in hierarchies, subtotals and so forth, and users will need to understand the concept of multiple views and datatypes, when to drop into EQL and so on – but for non-technical users working in an organization’s big data team it’s a great way to put a visual front-end onto the data in the data reservoir without having to understand tools like R Studio.

So that’s it for this three-part overview of Oracle Big Data Discovery and how it works with the Hadoop-based data reservoir. Keep an eye on the blog over the next few weeks as we get to grips with this new tool, and we’ll be covering it as part of the optional masterclass at the Brighton and Atlanta Rittman Mead BI Forum 2015 events this May.

Categories: BI & Warehousing

Why and How to use Oracle Metadata Management 12c. Part 2: Importing and Relating Metadata

Thu, 2015-02-26 05:27

In the first post of this series we have seen how to install and configure OEEM to start working with it. In this new post we are going to see how we import the metadata into OEMM from different sources like Oracle Database, OBIEE, ODI and OWB and then relate all of them inside OEMM.


After we have installed and configured OEMM, we need to start adding all the metadata from the different sources and applications that we use. In this example the sources will be some Oracle schemas and our applications will be ODI, OWB and OBIEE. To import the metadata for all of them we need to create one model in OEMM for each. A model in OEMM has all the connection details for a specific source or metadata provider (i.e: database schema, ODI repository, etc), and is also the container for the metadata of that specific source after the import process. So one model can connect to one specific source or application.

First and for organisational purposes we will create a Folder to contain the future models.  You can also create your models first, and then create the folder/s that you want and then just move the models under the correspondent folders. In addition, you can create folders within another folder.

To create a folder, right-click on the Repository entry under the Repository panel in the OEMM main page. Select New > Folder in the pop-up menu, enter a name and press the Create button.



The next step is creating the models and import the metadata of the different sources. The import or reverse engineering process is named harvesting in OEMM. We will start with the model for the Oracle Database. In this particular example I used Oracle 12c.

To create a model right click on the folder or the repository entry, and select New > Model. In the Create Model window that appears, enter a name for the new model and select the type of source that you want to import or to be more precise which will be the Import bridge that this model will use.

The Import Bridge is part of the Meta Integration® Model Bridge (MIMB) software and is the way that OEMM connect to the sources and applications to reverse engineering the metadata. You will find import bridges for a wide range of technologies like different databases, Business Intelligence  and Data Integration products from different vendors, Big Data stores, etc.


For this first example we will select the Oracle Database (via JDBC) import bridge and in the Import Setup tab we will add all the usual connection details: host, port, service and user and password to connect to the Database. This user should have at least the CONNECT privilege and the SELEC_CATALOG_ROLE role. We can also define this model for specific schemas using the magnifying glass to choose the shown schemas or just write the schemas (in uppercase) separated by “;”. Also we can decide if we want that the stored procedures are going to be included in this imported metadata or not.


After all the connection details have set, we test the connection and wait until we receive the Connection Successful message, and finally press the Create button. A message windows will appear asking if we want to “Import a new version now?” Press yes to start the harvesting process. A log window will show you the progress in the import process that can take several minutes. After the process is finished a new windows message ask if we want to open the model.


Choose yes to see all the objects that are imported for this model as it is shown in the figure below.


We need to repeat the process explained above to create the models for the rest of sources and applications that we are going to use in this example. The process is the same for all of them but of course there are some differences in the connection details required after we chose the specific Import Bridge for each one.

In the next screenshot you will find the connection details for the ODI model after you choose the Oracle Data Integrator (ODI) Import Bridge. In the Import Setup tab, you need to select the appropriate driver to connect to the database where is the ODI Repository (that could be Oracle, SQLServer, DB2, etc), the ODI Home folder, the URL to connect to database, the schema and the password for the Master Repository, user and password for the ODI User (SUPERVISOR for example), the name of the Work Repository from that we want to select the ODI Objects and the Context.


We need to select the Scope for this model between two options: Projects, that will include packages and mappings (or interfaces for versions before 12c) or Load Plans and Scenarios, that includes the Load Plans and Scenarios.

After we chose the Scope we can also filter the Content of the scope pressing the magnifying glass icon and select the specific objects that we want for this model.

After you press the create button to start the harvesting process, open the model created and it will look similar to this if you choosing Projects as Scope.


For the connection details to create the OWB model , you need to take a couple of things into account. First, the version of the OWB from which you want to import the metadata. If it is or later you will need to do the these two steps before:

  1. Copy the following .JAR files from: MetaIntegrationInstallationDir\java\ to %OWB_HOME%\owb\lib\ext\
  • jsr173_1.0_api.jar
  • MIR.jar
  • MIRModelBridge.jar
  • MIROracleWarehouseBuilderOmb.jar
  • MIRUtil.jar
  • stax-1.1.1-dev.jar
  1. Copy the MetaIntegrationInstallationDir\bin\mimbexec.bat file into the same OWB directory.

As the version that I have is, I copy the files detailed above, set the connection parameters like is shown in the following image and test the connection.

owb model config

When I started the import process the following error message appears in the log windows:



After trying many things unsuccessfully, I asked David Allan for help and he sent me another mimbexec.bat because apparently between and there were directory name changes.  This a temporary fix and a proper one is being worked on.

I substituted the bat file and I received another error message as it is shown in the next screenshot.



After a while, I realised that the issue that OEMM reported was because I was using an external table in one of the mappings. I changed it for a common table and the import process worked well. This has reported as a bug and a solution is being worked on. I really want to thank David for all his invaluable help on that.

This is how it looks the OWB model after the import of the metadata.


The last model that we need to create is the one based on OBIEE. There are different import bridges depending on the metadata that we need to import from OBIEE. Could be Oracle Business Intelligence (OBI) Server, Oracle Business Intelligence (OBI) Enterprise Edition and Oracle Business Intelligence (OBI) Answers.

The OBI Server import bridge needs the OBI repository in xml format as a parameter to import it, and the result model will contain all the objects defined in the three layers of the repository (Presentation, Business Model, Physical) as well as the repository connections, the variables and the initialisation blocks defined in the OBI repository.


To use the OBI Enterprise Edition import bridge we need to set the login user and password to connect to OBIEE (usually weblogic or a user with admin privileges), the repository file in xml format, and we can also filter the amount of reports retrieved from the OBI Presentation Server.

There are a couple of interesting not mandatory options, one is for optimise the import of large models which if it sets to true doesn’t return some objects like joins, relationships, logical fk, etc., to consume less memory at run time. And another option is to set if we want to do an incremental import to import only the changes of the source or each time we want to import everything.


The last import bridge to use with OBI is the OBI Answers, which will be import the content for a particular analysis or KPI report. This bridge needs to have the specific analysis in XML format.



About models, there are a couple of additional things that you need to take note. First if you want to see the configuration details you need to right-click the model and choose the settings option from the pop-up menu. In case that you want to open the model to see the objects that contains, double-click on it.

Another thing is for every parameter that you have in a model, you will find a very detailed help at the right in the import setup tab; and if you click on the name of the Import Bridge in the same tab, you have the documentation of this particular bridge which I find it very useful.

There are two tabs more in the folder and model definition that we won’t use in this example but that we talk in future posts: one for security and another to executing scripts when an event happens to this object. Models also have an additional tab Import Schedule, to create a plan to do the harvest process.

Relate the models

Once we have defined our models we need to relate them and to validate their relationship. The automated process of relate these models through the validation is named stitching. In order to do that we must create a Configuration first. A configuration in OEMM is a collection of models and another objects like mappings, glossaries, etc, that are related in someway.

According to the online documentation we need to consider a configuration as any of these options:

  • Repository workspace: a collection of Repository Objects to be analyzed together (search, browse, reports, etc.) as a technical scope, or business area under the same access permission scope.
  • Enterprise architecture – a collection of data store Models (ODS, data staging areas, data warehouses, data marts, etc.) and data process Models (ETL/DI, and BI) connected together through data flow stitching.
  • Design workflow – a collection of conceptual, logical and physical Models connected (semantically stitched) together through semantic mappings modeling the design process.

To create a Configuration, just right-click on a selected folder or the repository entry and choose New> Configuration. Enter a name for the configuration and press the Create button.


The configuration is opened and you need to drag the models that you want to be stitched inside this configuration as it is shown in the following screenshot


As you drag and drop your models, you can see that some of them have a warning icon after you include them in the configuration, and that is because we need to connect that model with the appropriate source of data.

To do that, select the model in the configuration and press Edit Connection. Choose the correspondent store for each connection and press OK.





















After you finish with all the models, press the Validate button, to start stitching or relate them.


In most of the cases, OEMM can assign the correspondent default schema for each of the connections in the model. If in some cases cannot do it , like in OWB, you need to do it manually.


In the following image you will see all the models validated. For this example, I’ve created four databases models, one that contains the source (transactional system), one for the staging schema, and another two that contains different data warehouses. Also an ODI model, an OWB model and an OBIEE model.


You can see also the relationship between the models that belong to a configuration in a graphical view in the Architecture Diagram tab of the Configuration. If the diagram looks like a little messy, you  can press the Edit button and then Layout to order the way the components are shown.


In summary, we create and harvesting (reverse-engineer) the models and then relate or stitching them to can analyse them together. In the next post, we will see some interesting stuff that we can do with configurations and models like trace data lineage and trace data impact.

Categories: BI & Warehousing

Introducing Oracle Big Data Discovery Part 2: Data Transformation, Wrangling and Exploration

Wed, 2015-02-25 09:51

In yesterday’s post I looked at Oracle Big Data Discovery and how it brought the search and analytic capabilities of Endeca to Hadoop. We looked at how the Oracle Endeca Information Discovery Studio application works with a version of the Endeca Server engine to analyse and visualise sample sets of data from the Hadoop cluster, and how it uses Apache Spark to retrieve data from Hadoop and then transform that data to make it more suitable for data discovery and data analysis applications. Oracle Big Data Discovery is designed to work alongside ODI and GoldenGate for Big Data once you’ve decided on your main data flows, and Oracle Big Data SQL for BI tool and application access to the entire “data reservoir”. So how does Big Data Discovery work, and what role does it play in the overall big data project workflow?

The best way to think of Big Data Discovery, to my mind, is “Endeca on Hadoop”. Endeca Information Discovery had three main parts to it; the data loading part performed using Endeca Information Discovery Integrator and more recently, the personal data upload feature in Endeca Information Discovery Studio. Data was then ingested into the Endeca Server engine and stored in a key/value-store NoSQL database, indexed, parsed and enriched, and then analyzed using the graphical user interface provided by Studio. As I explained in more detail in my first post in the series yesterday, Big Data Discovery runs the Studio and DGraph (Endeca Server) elements on one or more dedicated nodes, and then reads data in from Hadoop and then writes it back in transformed states using Apache Spark, as shown in the diagram below:


As the data discovery and analysis features in Big Data Discovery rely on getting data into the DGraph (Endeca Server) engine first of all, this implies two things; first, we’ll need to take a subset or sample of the entire Hadoop dataset and load just that into the DGraph engine, and second we’ll need some means of transforming and “massaging” that data so it works well as a data discovery set, and then writing those changes back to the full Hadoop dataset if we want to use it with some other tool – OBIEE or Big Data SQL, for example. To see how this process works, let’s use the same Rittman Mead Apache webserver logs that I’ve used in my previous examples, and bring that data and some additional reference data into Big Data Discovery.

The log data from the RM webserver is in Apache Combined Log Format and a sample of the rows looks like this:


For data to be eligible to be ingested into Big Data Discovery, it has to be registered in the Hive Metastore and with the metadata available to use by external tools using the HCatalog service. This means that you already need to have created a Hive table over each datasource, either pointing this table to regular fixed-width or delimited files, or using a SerDe to translate another file format – say a compressed/column-store format like Parquet – into a format that Hive can understand. In our case I can use the RegEx SerDe that I first used in this blog post a while ago to create a Hive table over the log file and split out the various log file elements, with the resulting DDL looking like this:

host STRING,
identity STRING,
user STRING,
time STRING,
request STRING,
status STRING,
size STRING,
referer STRING,
agent STRING)
ROW FORMAT SERDE 'org.apache.hadoop.hive.contrib.serde2.RegexSerDe'
"input.regex" = "([^ ]*) ([^ ]*) ([^ ]*) (-|\\[[^\\]]*\\]) 
([^ \"]*|\"[^\"]*\") (-|[0-9]*) (-|[0-9]*)(?: ([^ \"]*|\"
[^\"]*\") ([^ \"]*|\"[^\"]*\"))?",
"output.format.string" = "%1$s %2$s %3$s %4$s %5$s %6$s %7$s %8$s %9$s"
LOCATION '/user/oracle/rm_logs';

If I then register the SerDe with Big Data Discovery I could ingest the table and file at this point, or I can use a Hive CTAS statement to remove the dependency on the SerDe and ingest into BDD without any further configuration.

create table access_logs as
select * 
from apachelog;

At this point, if you’ve got the BDD Hive Table Detector running, it should pick up the presence of the new hive table and ingest it into BDD (you can whitelist table names, and restrict it to certain Hive databases if needed). Or, you can manually trigger the ingestion from the Data Processing CLI on the BDD node, like this:

[oracle@bddnode1 ~]$ cd /home/oracle/Middleware/BDD1.0/dataprocessing/edp_cli
[oracle@bddnode1 edp_cli]$ ./data_processing_CLI -t access_logs;

The data processing process then creates an Apache Oozie job to sample a statistically relevant sample set of data into Apache Spark – with a 1% sample providing 95% sample accuracy – that is the profiled, enriched and then loaded into the Big Data Discovery DGraph engine for further transformation, then exploration and analysis within Big Data Discovery Studio.


The profiling step in this process scans the incoming data and helps BDD determine the datatype of each Hive table column, the distribution of values within the column and so on, whilst the enrichment part identifies key words and phrases and other key lexical facts about the dataset. A key concept here also is that BDD typically works with a representative sample of your Hive table contents, not the whole contents, as all the data you analyse has to fit within the memory space with the DGraph engine, just like it used to with Endeca Server. At some point its likely that the functionality of the DGraph engine will be unbundled from the Endeca Server and run natively across the actual Hadoop cluster, but for now you have to separately ingest data into the DGraph engine (which can run clustered on BDD nodes) and analyse it there – however the rules of sampling are that if you’ve got a sufficiently big sample – say, 1m rows – regardless of the actual main dataset size this sample set is considered sufficiently representative – 95% in this case – as to make loading a bigger sample set not really worth the effort. But bear in mind when working with a BDD dataset that you’re working a sample, not the full set, so if a value you’re looking for is missing it might be because it’s not in this particular sample.

Once you’ve ingested the new dataset into BDD, you see it listed amongst the others that have previously been ingested, like this:


At this point you can explore the dataset, to take an initial look at the patterns and values in the dataset in its raw form.


Unfortunately, in this raw form the data in the access_logs table isn’t all that useful – details of the page request URL are mixed in with the HTTP protocol and method, for example; dates are in strings; details of the person accession the site are in IP address format rather than a geographical location, and so on. In previous examples on this blog I’ve looked at various methods to cleanse, transform and enhance the data in log file tables like this, using tools and techniques such as Hive table transformations, Pig and Apache Spark scripts, and ODI mappings but all of these typically require some IT invovement whereas one of the hallmarks of recent versions of Endeca Information Discovery Studio was giving power-users the ability to transform and enrich data themselves. Big Data Discovery provides tools to cleanse, transform and enrich data, with menu items for common transformations and a Groovy script editor for more complex ones, including deriving sentiment values from textual data and stripping out HTML and formatting characters from text.


Once you’ve finished transforming and enriching the dataset, you can either save (commit) the changes back to the sample dataset in the BDD DGraph engine, or you can use the transformation rules you’ve defined to apply those transformations to the entire Hive table contents back on Hadoop, with the transformation work being done using Apache Spark. Datasets are loaded into “projects” and each project can have its own transformed view of the raw data, with copies of the dataset being kept in the BDD DGraph engine to represent each team’s specific view onto the raw datasets.


In practice I found this didn’t, at the current product state, completely replace the need for a Hadoop developer or R data analyst – you need to get your data files into Hive and HCatalog at the start which involves parsing and interpreting semi-structured data files, and I often did some transformations in BDD, then applied the transformations to the whole Hive dataset and then re-imported the results back into BDD to start from a simple known state. But it certainly made tasks such as turning IP addresses into countries and cities, splitting our URLs and removing HTML tags much easier and I got the data cleansing process done in a matter of hours compared to the days with manual Hive, Pig and Spark scripting.

Now the data in my log file dataset is much more usable and easy to understand, with URLs split out, status codes grouped into high-level descriptors, and other descriptive and formatting changes made.


I can also at this point bring in additional datasets, either created manually outside of BDD and ingested into the DGraph from Hive, or manually uploaded using the Studio interface. These dataset uploads then live in the BDD DGraph engine, and are then written back to Hive for long-term persistence or for sharing with other tools and processes.


These datasets can then be joined to the main dataset on matching dataset columns, giving you a table-join interface not unlike OBIEE’s physical model editor.


So now we’re in a position where our datasets have been ingested into BDD, and we’ve cleansed, transformed and joined them into a combined web activity dataset. In tomorrow’s final post I’ll look at the data visualisation part of Big Data Discovery and see how it brings the capabilities of Endeca Information Discovery Studio to Hadoop.

Categories: BI & Warehousing

Introducing Oracle Big Data Discovery Part 1: “The Visual Face of Hadoop”

Mon, 2015-02-23 20:38

Oracle Big Data Discovery was released last week, the latest addition to Oracle’s big data tools suite that includes Oracle Big Data SQL, ODI and it’s Hadoop capabilities and Oracle GoldenGate for Big Data 12c. Introduced by Oracle as “the visual face of Hadoop”, Big Data Discovery combines the data discovery and visualisation elements of Oracle Endeca Information Discovery with data loading and transformation features built on Apache Spark to deliver a tool aimed at the “Discovery Lab” part of the Oracle Big Data and Information Management Reference Architecture.


Most readers of this blog will probably be aware of Oracle Endeca Information Discovery, based on the Endeca Latitude product acquired as part of the Endeca aquisition. Oracle positioned Endeca Information Discovery (OEID) in two main ways; on the one hand as a data discovery tool for textual and unstructured data that complemented the more structured analysis capabilities of Oracle Business Intellligence, and on the other hand, as a fast click-and-refine data exploration tool similar to Qlikview and Tableau.

The problem for Oracle though was that data discovery against files and documents is a bit of a “solution looking for a problem” and doesn’t have a naturally huge market (especially considering the license cost of OEID Studio and the Endeca Server engine that stores and analyzes the data), whereas Qlikview and Tableau are significantly cheaper than OEID (at least at the start) and are more focused on BI-type tasks, making OEID a good too but not one with a mass market. To address this, whilst OEID will continue as a standalone tool the data discovery and unstructured data analysis parts of OEID are making their way into this new product called Oracle Big Data Discovery, whilst the fast click-and-refine features will surface as part of Visual Analyzer in OBIEE12c.

More importantly, Big Data Discovery will run on Hadoop making it a solution for a real problem – how to catalog, explore, refine and visualise the data in the data reservoir, where data has been landed that might be in schema-on-read databases, might need further analysis and understanding, and users need large-scale tooling to extract the nuggets of information that in time make their way into the “Execution” part of the Big Data and Information Management Reference Architecture. As some who’s admired the technology behind Endeca Information Discovery but sometimes struggled to find real-life use-cases or customers for it, I’m really pleased to see its core technology applied to a problem space that I’m encountering every day with Rittman Mead’s customers.


In this first post, I’ll look at how Big Data Discovery is architected and how it works with Cloudera CDH5, the Hadoop distribution we use with our customers (Hortonworks HDP support is coming soon). In the next post I’ll look at how data is loaded into Big Data Discovery and then cataloged and transformed using the BDD front-end; then finally, we’ll take a look at exploring and analysing data using the visual capabilities of BDD evolved from the Studio tool within OEID. Oracle Big Data Discovery 1.0 is now GA (Generally Available) but as you’ll see in a moment you do need a fairly powerful setup to run it, at least until such time as Oracle release a compact install version running on VM.

To run Big Data Discovery you’ll need access to a Hadoop install, which in most cases will consist of 6 (minumum 3 or 4, but 6 is the minimum we use) to 18 or so Hadoop nodes running Cloudera CDH5.3. BDD generally runs on its own server nodes and itself can be clustered, but for our setup we ran 1 BDD node alongside 6 CDH5.3 Hadoop nodes looking like this:


Oracle Big Data Discovery is made up of three component types highlighted in red in the above diagram, two of which typically run on their own dedicated BDD nodes and another which runs on each node in the Hadoop cluster (though there are various install types including all on one node, for demo purposes)

  • The Studio web user interface, which combines the faceted search and data discovery parts of Endeca Information Discovery Studio with a lightweight data transformation capability
  • The DGraph Gateway, which brings Endeca Server search/analytics capabilities to the world of Hadoop, and
  • The Data Processing component that runs on each of the Hadoop nodes, and uses Hive’s HCatalog feature to read Hive table metadata and Apache Spark to load and transform data in the cluster

The Studio component can run across several nodes for high-availability and load-balancing, which the DGraph element can run on a single node as I’ve set it up, or in a cluster with a single “leader” node and multiple “follower” nodes again for enhanced availability and throughput. The DGraph part them works alongside Apache Spark to run intensive search and analytics on subsets of the whole Hadoop dataset, with sample sets of data being moved into the DGraph engine and any resulting transformations then being applied to the whole Hadoop dataset using Apache Spark. All of this then runs as part of the wider Oracle Big Data product architecture, which uses Big Data Discovery and Oracle R for the discovery lab and Oracle Exadata, Oracle Big Data Appliance and Oracle Big Data SQL to take discovery lab innovations to the wider enterprise audience.


So how does Oracle Big Data Discovery work in practice, and what’s a typical workflow? How does it give us the capability to make sense of structured, semi-structured and unstructured data in the Hadoop data reservoir, and how does it look from the perspective of an Oracle Endeca Information Discovery developer, or an OBIEE/ODI developer? Check back for the next parts in this three part series where I’ll first look at the data transformation and exploration capabilities of Big Data Discovery, and then look at how the Studio web interface brings data discovery and data visualisation to Hadoop.

Categories: BI & Warehousing

Data Integration Tips: ODI – Use Scenarios in packages

Tue, 2015-02-10 07:16


Here is another question I often hear during ODI Bootcamp or I often read in the ODI Space on OTN :
In a package, should I directly use mappings (or interfaces prior to 12c) or is it better to use the scenarios generated from these mappings?

An ODI package is a workflow used to sequence the different steps of execution of a data integration process. Some of these steps might change the value of a variable, evaluate a variable, perform an administrative task like sending an email or retrieving a file from an FTP. But the core thing is the execution of a mapping and this can be done in two different ways.


Direct Use

It is possible to directly drag and drop a mapping in the package.

Mapping in a Package

In this example the mapping M_FACT_SALES_10 is directly executed within the package, in the same session. This means that the value of the variables will be the same as the ones in the package.

Execution of a package holding a mapping


If there are several mappings, they will be executed sequentially, there is no parallelism here. (Damn, I will never get the double L at the right place in that word… Thanks autocorrect!)

The code for the mapping is generated using the current state of the mapping in the work repository, including any changes done since the package was created. But if a scenario is created for the package PCK_FACT_LOAD_MAP, it will need to be regenerated – or a new version should be generated – to take any change of M_FACT_SALES_10 into account.

One good point for this approach is that we can build a data integration process that will be entirely rolled back in case of failure. If one of the step fails, the data will be exactly as it was before the execution started. This can be done by using the same transaction for all the mappings and disable the commit option for all of them but the last one. So it’s only when everything succeeds that the changes will be committed and available for everyone. Actually it’s not even required to commit on the last step as ODI will anyway issue a commit at the end of a successful session. This technique works of course only if you don’t have any DDL statement on the target tables. You will find more information about it on this nice blog post from Bhabani Rajan : Transaction Control in ODI. Thanks to David Allan for reminding me this!


Using Scenarios

But you can also generate a scenario and then drag and drop it in a packageScenario in a package

Sessions and variables

In this case, the execution of the mapping will not be done within the package session. Instead, on step of this session will use the OdiStartScen command to trigger a new session to execute the scenario.

Execution of a package holding a scenario

We can see here that the second step of the package session (401) got only one task which will run the command to start a new session. The only step in the new session (402) is the mapping and it has the three same tasks as the previous example. Thanks to the fact this is a different session, you could choose to execute it using another context or agent. My colleague Michael brought an interesting use case to me. When one step of your package must extract data from a file on a file server that has its own standalone agent installed to avoid permission issues, you could specify that agent in the Scenario step of your package. So the session for that scenario will be started on that standalone agent while all the other sessions will use the execution agent of the parent session.


But what about variables configured to keep no history? As this is a different session, the values are lost outside of the scope of one session. We therefore need to pass it as startup parameters to the scenario of the mapping.

Passing variables as startup parameters To do this, I go to the Command tab on the scenario step and I add my variables there with the syntax


where <VALUE_TO_ASSIGN> can be the value of variable in the current session itself (401 in our case).

 Code executed

The scenario step name was originally “Execution of the Scenario M_FACT_SALES_10 version 001″ and I slightly modified it to remove the mention of a version number. As I want to always execute the last version of the scenario, I also changed the version in the parameters and set it to -1 instead of 001. There is a small bug in the user interface so if you click somewhere else, it will change back to 001. The workaround is to press the Return key (or Enter) after typing your value.

So by using a scenario, you can choose which version of a scenario you want to execute. Either a specific version (e.g. 003) or the last one, using the value -1. It also means that if you won’t break anything if the package is executed while you are working on the mapping. The package will still use the frozen code of the scenario even if you changed the mapping.

If you create a scenario for the package PCK_FACT_LOAD_SCEN, there is no need to regenerate it when a new scenario is (re-)generated for the mapping. By using -1, it will reference the newly generated scenario.

 Asynchronous Execution

Another advantage of using scenarios is that it supports parallelism (Yes, I got it right!!).

Asynchronous scenario executions

Here I set the “Synchronous / Asynchronous” option to “Asynchronous Mode” for my two scenarios so the two sessions will start at the same time. By adding an OdiWaitForChildSessions step, I can wait for the end of all the sessions before doing anything else. This step will also define in which case you want to report an error. By default, it’s when all the child sessions are in error. But I usually change to parameter to zero, so any failure will cause my package execution to fail as well.


Just a short note, be careful when using the method getPrevStepLog() from the substitution API in a step after executing a scenario. That method will retrieve information about the previous step in the current Session. So about the ODI command execution OdiStartScen, and not about the execution of the scenario itself.



Here is a small recap table to compare both approach :

comparison table

In conclusion, development is generally more robust when using scenarios in packages. There is more control on what is executed and on the parallelism. The only good reason to use mappings directly is to keep everything within the same transaction.

Keep also in mind that Load Plans might be a better alternative than packages with better parallelism, error handling and restartability. Unless you need loops or a persistent session… A comparison between Load Plans and Packages can be found in the Oracle blog post introducing Load Plans.


More tips and core principles demystification coming soon so stay tuned. And follow @mRainey, @markrittman and myself – @JeromeFr – on twitter to be informed of anything happening in the ODI world.

Categories: BI & Warehousing

OBIEE Monitoring and Diagnostics with InfluxDB and Grafana

Sun, 2015-02-08 15:21

In this article I’m going to look at collecting time-series metrics into the InfluxDB database and visualising them in snazzy Grafana dashboards. The datasets I’m going to use are OS metrics (CPU, Disk, etc) and the DMS metrics from OBIEE, both of which are collected using the support for a Carbon/Graphite listener in InfluxDB.

The Dynamic Monitoring System (DMS) in OBIEE is one of the best ways of being able to peer into the internals of the product and find out quite what’s going on. Whether performing diagnostics on a specific issue or just generally monitoring to make sure things are ticking over nicely, using the DMS metrics you can level-up your OBIEE sysadmin skills beyond what you’d get with Fusion Middleware Control out of the box. In fact, the DMS metrics are what you can get access to with Cloud Control 12c (EM12c) – but for that you need EM12c and the BI Management Pack. In this article we’re going to see how to easily set up our DMS dashboard.
N.B. if you’ve read my previous articles, what I write here (use InfluxDB/Grafana) supersedes what I wrote in those (use Graphite) as my recommended approach to working with arbitrary time-series metrics.


To get the DMS data out of OBIEE we’re going to use the obi-metrics-agent tool that Rittman Mead open-sourced last year. This connects to OPMN and pulls the data out. We’ll store the data in InfluxDB, and then visualise it in Grafana. Whilst not mandatory for the DMS stats, we’ll also setup collectl so that we can show OS stats alongside the DMS ones.


InfluxDB is a database, but unlike a RDBMS such as Oracle – good for generally everything – it is a what’s called a Time-Series Database (TSDB). This category of database focuses on storing data for a series, holding a given value for a point in time. Generally they’re optimised for handling large quantities of inbound metrics (think Internet of Things), rather than necessarily excelling at handling changes to the data (update/delete) – but that’s fine here since metric events in the past don’t generally change.

I’m using InfluxDB here for a few reasons:

  1. Grafana supports it as a source, with lots of active development for its specific features.
  2. It’s not Graphite. Whilst I have spent many a happy hour using Graphite I’ve spent many a frustrating day and night trying to install the damn thing – every time I want to use it on a new installation. It’s fundamentally long in the tooth, a whilst good for its time is now legacy in my mind. Graphite is also several things – a data store (whisper), a web application (graphite web), and a data collector (carbon). Since we’re using Grafana, the web front end that Graphite provides is redundant, and is where a lot of the installation problems come from.
  3. KISS! Yes I could store time series data in Oracle/mySQL/DB2/yadayada, but InfluxDB does one thing (storing time series metrics) and one thing only, very well and very easily with almost no setup.

For an eloquent discussion of Time-Series Databases read these couple of excellent articles by Baron Schwarz here and here.


On the front-end we have Grafana which is a web application that is rapidly becoming accepted as one of the best time-series metric visualisation tools available. It is a fork of Kibana, and can work with data held in a variety of sources including Graphite and InfluxDB. To run Grafana you need to have a web server in place – I’m using Apache just because it’s familiar, but Grafana probably works with whatever your favourite is too.


This article is based around the OBIEE SampleApp v406 VM, but should work without modification on any OL/CentOS/RHEL 6 environment.

InfluxDB and Grafana run on both RHEL and Debian based Linux distros, as well as Mac OS. The specific setup steps detailed here might need some changes according on the OS.

Getting Started with InfluxDB InfluxDB Installation and Configuration as a Graphite/Carbon Endpoint

InfluxDB is a doddle to install. Simply download the rpm, unzip it, and run. BOOM. Compared to Graphite, this makes it a massive winner already.

sudo rpm -ivh influxdb-latest-1.x86_64.rpm

This downloads and installs InfluxDB into /opt/influxdb and configures it as a service that will start at boot time.

Before we go ahead an start it, let’s configure it to work with existing applications that are sending data to Graphite using the Carbon protocol. InfluxDB can support this and enables you to literally switch Graphite out in favour of InfluxDB with no changes required on the source.

Edit the configuration file that you’ll find at /opt/influxdb/shared/config.toml and locate the line that reads:


In v0.8.8 this is at line 41. In the following stanza set the plugin to enabled, specify the listener port, and give the name of the database that you want to store data in, so that it looks like this.

# Configure the graphite api
enabled = true
# address = "" # If not set, is actually set to bind-address.
port = 2003
database = "carbon"  # store graphite data in this database
# udp_enabled = true # enable udp interface on the same port as the tcp interface

Note that the file is owned by a user created at installation time, influxdb, so you’ll need to use sudo to edit the file.

Now start up InfluxDB:

sudo service influxdb start

You should see it start up successfully:

[oracle@demo influxdb]$ sudo service influxdb start
Setting ulimit -n 65536
Starting the process influxdb [ OK ]
influxdb process was started [ OK ]

You can see the InfluxDB log file and confirm that the Graphite/Carbon listener has started:

[oracle@demo shared]$ tail -f /opt/influxdb/shared/log.txt
[2015/02/02 20:24:04 GMT] [INFO] (·005:1187) Recovered local server
[2015/02/02 20:24:04 GMT] [INFO] (*Server).ListenAndServe:133) recovered
[2015/02/02 20:24:04 GMT] [INFO] (*Coordinator).ConnectToProtobufServers:898) Connecting to other nodes in the cluster
[2015/02/02 20:24:04 GMT] [INFO] (*Server).ListenAndServe:139) Starting admin interface on port 8083
[2015/02/02 20:24:04 GMT] [INFO] (*Server).ListenAndServe:152) Starting Graphite Listener on
[2015/02/02 20:24:04 GMT] [INFO] (*Server).ListenAndServe:178) Collectd input plugins is disabled
[2015/02/02 20:24:04 GMT] [INFO] (*Server).ListenAndServe:187) UDP server is disabled
[2015/02/02 20:24:04 GMT] [INFO] (*Server).ListenAndServe:187) UDP server is disabled
[2015/02/02 20:24:04 GMT] [INFO] (*Server).ListenAndServe:216) Starting Http Api server on port 8086
[2015/02/02 20:24:04 GMT] [INFO] (*Server).reportStats:254) Reporting stats: &client.Series{Name:"reports", Columns:[]string{"os", "arch", "id", "version"}, Points:[][]interface {}{[]interface {}{"linux", "amd64", "e7d3d5cf69a4faf2", "0.8.8"}}}

At this point if you’re using the stock SampleApp v406 image, or if indeed any machine with a firewall configured, you need to open up ports 8083 and 8086 for InfluxDB. Edit /etc/sysconfig/iptables (using sudo) and add:

-A INPUT -m state --state NEW -m tcp -p tcp --dport 8083 -j ACCEPT
-A INPUT -m state --state NEW -m tcp -p tcp --dport 8086 -j ACCEPT

immediately after the existing ACCEPT rules. Restart iptables to pick up the change:

sudo service iptables restart

If you now go to http://localhost:8083/ (replace localhost with the hostname of the server on which you’ve installed InfluxDB), you’ll get the InfluxDB web interface. It’s fairly rudimentary, but suffices just fine:

Login as root/root, and you’ll see a list of nothing much, since we’ve not got any databases yet. You can create a database from here, but for repeatability and a general preference for using the command line here is how to create a database called carbon with the HTTP API called from curl (assuming you’re running it locally; change localhost if not):

curl -X POST 'http://localhost:8086/db?u=root&p=root' -d '{"name": "carbon"}'

Simple huh? Now hit refresh on the web UI and after logging back in again you’ll see the new database:

You can call the database anything you want, just make sure what you create in InfluxDB matches what you put in the configuration file for the graphite/carbon listener.

Now we’ll create a second database that we’ll need later on to hold the internal dashboard definitions from Grafana:

curl -X POST 'http://localhost:8086/db?u=root&p=root' -d '{"name": "grafana"}'

You should now have two InfluxDB databases, primed and ready for data:

Validating the InfluxDB Carbon Listener

To make sure that InfluxDB is accepting data on the carbon listener use the NetCat (nc) utility to send some dummy data to it:

echo " 3 `date +%s`"|nc localhost 2003

Now go to the InfluxDB web interface and click Explore Data ». In the query field enter

list series

To see the first five rows of data itself use the query

select * from /.*/ limit 5

InfluxDB Queries

You’ll notice that what we’re doing here (“SELECT … FROM …”) looks pretty SQL-like. Indeed, InfluxDB support a SQL-like query language, which if you’re coming from an RDBMS background is nicely comforting ;-)

The syntax is documented, but what I would point out is the apparently odd /.*/ constructor for the “table” is in fact a regular expression (regex) to match the series for which to return values. We could have written select * from but the .* wildcard enclosed in the / / regex delimiters is a quick way to check all the series we’ve got.

Going off on a bit of a tangent (but hey, why not), let’s write a quick Python script to stick some randomised data into InfluxDB. Paste the following into a terminal window to create the script and make it executable:

cat >~/<<EOF
#!/usr/bin/env python
import socket
import time
import random
import sys

CARBON_SERVER = sys.argv[1]
CARBON_PORT = int(sys.argv[2])

while True:
        message = ' %d %d\n' % (random.randint(1,20),int(time.time()))
        print 'sending message:\n%s' % message
        sock = socket.socket()
        sock.connect((CARBON_SERVER, CARBON_PORT))
chmod u+x ~/

And run it: (hit Ctrl-C when you’ve had enough)

$ ~/ localhost 2003
sending message: 3 1422910401

sending message: 5 1422910402

Now we’ve got two series in InfluxDB:

  • – that we sent using nc
  • – using the python script

Let’s go back to the InfluxDB web UI and have a look at the new data, using the literal series name in the query:

select * from

Well fancy that – InfluxDB has done us a nice little graph of the data. But more to the point, we can see all the values in the series.

And a regex shows us both series, matching on the ‘foo’ part of the name:

select * from /foo/ limit 3

Let’s take it a step further. InfluxDB supports aggregate functions, such as max, min, and so on:

select count(value), max(value),mean(value),min(value) from

Whilst we’re at it, let’s bring in another way to get data out – with the HTTP API, just like we used for creating the database above. Given a query, it returns the data in json format. There’s a nice little utility called jq which we can use to pretty-print the json, so let’s install that first:

sudo yum install -y jq

and then call the InfluxDB API, piping the return into jq:

curl --silent --get 'http://localhost:8086/db/carbon/series?u=root&p=root' --data-urlencode "q=select count(value), max(value),mean(value),min(value) from"|jq '.'

The result should look something like this:

    "name": "",
    "columns": [
    "points": [

We could have used the Web UI for this, but to be honest the inclusion of the graphs just confuses things because there’s nothing to graph and the table of data that we want gets hidden lower down the page.

Setting up obi-metrics-agent to Send OBIEE DMS metrics to InfluxDB

obi-metrics-agent is an open-source tool from Rittman Mead that polls your OBIEE system to pull out all the lovely juicy DMS metrics from it. It can write them to file, insert them to an RDBMS, or as we’re using it here, send them to a carbon-compatible endpoint (such as Graphite, or in our case, InfluxDB).

To install it simply clone the git repository (I’m doing it to /opt but you can put it where you want)

# Install pre-requisite
sudo yum install -y libxml2-devel python-devel libxslt-devel python-pip
sudo pip install lxml
# Clone the git repository
git clone ~/obi-metrics-agent
# Move it to /opt folder
sudo mv ~/obi-metrics-agent /opt

and then run it:

cd /opt/obi-metrics-agent

./ \
--opmnbin /app/oracle/biee/instances/instance1/bin/opmnctl \
--output carbon \
--carbon-server localhost

I’ve used line continuation character \ here to make the statement clearer. Make sure you update opmnbin for the correct path of your OPMN binary as necessary, and localhost if your InfluxDB server is not local to where you are running obi-metrics-agent.

After running this you should be able to see the metrics in InfluxDB. For example:

select * from /Oracle_BI_DB_Connection_Pool\..+\.*Busy/ limit 5

Setting up collectl to Send OS metrics to InfluxDB

collectl is an excellent tool written by Mark Seger and reports on all sorts of OS-level metrics. It can run interactively, write metrics to file, and/or send them on to a carbon endpoint such as InfluxDB.

Installation is a piece of cake, using the EPEL yum repository:

# Install the EPEL yum repository
sudo rpm -Uvh`uname -p`/epel-release-6-8.noarch.rpm
# Install collectl
sudo yum install -y collectl
# Set it to start at boot
sudo chkconfig --level 35 collectl on

Configuration to enable logging to InfluxDB is a simple matter of modifying the /etc/collectl.conf configuration file either by hand or using this set of sed statements to do it automagically.
The localhost in the second sed command is the hostname of the server on which InfluxDB is running:

sudo sed -i.bak -e 's/^DaemonCommands/#DaemonCommands/g' /etc/collectl.conf
sudo sed -i -e '/^#DaemonCommands/a DaemonCommands = -f \/var\/log\/collectl -P -m -scdmnCDZ --export graphite,localhost:2003,p=.os,s=cdmnCDZ' /etc/collectl.conf

If you want to log more frequently than ten seconds, make this change (for 5 second intervals here):

sudo sed -i -e '/#Interval =     10/a Interval = 5' /etc/collectl.conf

Restart collectl for the changes to take effect:

sudo service collectl restart

As above, a quick check through the web UI should confirm we’re getting data through into InfluxDB:

Note the very handy regex lets us be lazy with the series naming. We know there is a metric called in part ‘cputotal’, so using /cputotal/ can match anything with it in.

Installing and Configuring Grafana

Like InfluxDB, Grafana is also easy to install, although it does require a bit of setting up. It needs to be hooked into a web server, as well as configured to connect to a source for metrics and storing dashboard definitions.

First, download the binary (this is based on v1.9.1, but releases are frequent so check the downloads page for the latest):

cd ~

Unzip it and move it to /opt:

sudo mv grafana-1.9.1 /opt

Configuring Grafana to Connect to InfluxDB

We need to do a bit of configuration, so first create the configuration file based on the template given:

cd /opt/grafana-1.9.1
cp config.sample.js config.js

And now open the config.js file in your favourite text editor. Grafana supports various sources for metrics data, as well as various targets to which it can save the dashboard definitions. The configuration file helpfully comes with configuration elements for many of these, but all commented out. Uncomment the InfluxDB stanzas and amend them as follows:

datasources: {
    influxdb: {
        type: 'influxdb',
        url: "http://sampleapp:8086/db/carbon",
        username: 'root',
        password: 'root',
    grafana: {
        type: 'influxdb',
        url: "http://sampleapp:8086/db/grafana",
        username: 'root',
        password: 'root',
        grafanaDB: true

Points to note:

  1. The servername is the server host as you will be accessing it from your web browser. So whilst the configuration we did earlier was all based around ‘localhost’, since it was just communication within components on the same server, the Grafana configuration is what the web application from your web browser uses. So unless you are using a web browser on the same machine as where InfluxDB is running, you must put in the server address of your InfluxDB machine here.
  2. The default InfluxDB username/password is root/root, not admin/admin
  3. Edit the database names in the url, either as shown if you’ve followed the same names used earlier in the article or your own versions of them if not.
Setting Grafana up in Apache

Grafana runs within a web server, such as Apache or nginx. Here I’m using Apache, so first off install it:

sudo yum install -y httpd

And then set up an entry for Grafana in the configuration folder by pasting the following to the command line:

cat > /tmp/grafana.conf <<EOF
Alias /grafana /opt/grafana-1.9.1

<Location /grafana>
Order deny,allow
Allow from
Allow from ::1
Allow from all

sudo mv /tmp/grafana.conf /etc/httpd/conf.d/grafana.conf

Now restart Apache:

sudo service httpd restart

And if the gods of bits and bytes are smiling on you, when you go to http://yourserver/grafana you should see:

Note that as with InfluxDB, you may well need to open your firewall for Apache which is on port 80 by default. Follow the same iptables instructions as above to do this.

Building Grafana Dashboards on Metrics Held in InfluxDB

So now we’ve set up our metric collectors, sending data into InfluxDB.

Let’s see now how to produce some swanky dashboards in Grafana.

Grafana has a concept of Dashboards, which are made up of Rows and within those Panels. A Panel can have on it a metric Graphs (duh), but also static text or single figure metrics.

To create a new dashboard click the folder icon and select New:

You get a fairly minimal blank dashboards. On the left you’ll notice a little green tab: hover over that and it pops out to form a menu box, from where you can choose the option to add a graph panel:

Grafana Graph Basics

On the blank graph that’s created click on the title (with the accurate text “click here”) and select edit from the options that appear. This takes you to the graph editing page, which looks equally blank but from here we can now start adding metrics:

In the box labelled series start typing Active_Sessions and notice that Grafana will autocomplete it to any available metrics matching this:

Select Oracle_BI_PS_Sessions.Active_Sessions and your graph should now display the metric.

To change the time period shown in the graph, use the time picker at the top of the screen.You can also click & drag (“brushing”) on any graph to select a particular slice of time.

So, set the time filter to 15 minutes ago and from the Auto-refresh submenu set it to refresh every 5 seconds. Now login to your OBIEE instance, and you should see the Active Sessions value increase (one per session login):

To add another to the graph you can click on Add query at the bottom right of the page, or if it’s closely related to the one you’ve defined already click on the cog next to it and select duplicate:

In the second query add Oracle_BI_General.Total_sessions (remember, you can just type part of the string and Grafana autocompletes based on the metric series stored in InfluxDB). Run a query in OBIEE to cause sessions to be created on the BI Server, and you should now see the Total sessions increase:

To save the graph, and the dashboard, click the Save icon. To return to the dashboard to see how your graph looks alongside others, or to add a new dashboards, click on Back to dashboard.

Grafana Graph Formatting

Let’s now take a look at the options we’ve got for modifying the styling of the graph. There are several tabs/sections to the graph editor – General, Metrics (the default), Axes & Grid, and Display Styles. The first obvious thing to change is the graph title, which can be changed on the General tab:

From here you can also change how the graph is sized on the dashboard using the Span and Height options. A new feature in recent versions of Grafana is the ability to link dashboards to help with analysis paths – guided navigation as we’d call it in OBIEE – and it’s from the General tab here that you can define this.

On the Metrics tab you can specify what text to use in the legend. By default you get the full series name, which is usually too big to be useful as well as containing a lot of redundant repeating text. You can either specify literal text in the alias field, or you can use segments of the series name identified by $x where x is the zero-based segment number. In the example I’ve hardcoded the literal value for the second metric query, and used a dynamic segment name for the first:

On the Axes & Grid tab you can specify the obvious stuff like min/max scales for the axes and the scale to use (bits, bytes, etc). To put metrics on the right axis (and to change the colour of the metric line too) click on the legend line, and from there select the axis/colour as required:

You can set thresholds to overlay on the graph (to highlight warning/critical values, for example), as well as customise the legend to show an aggregate value for each metric, show it in a table, or not at all:

The last tab, Display Styles, has even more goodies. One of my favourite new additions to Grafana is the Tooltip. Enabling this gives you a tooltip when you hover over the graph, displaying the value of all the series at that point in time:

You can change the presentation of the graph, which by default is a line, adding bars and/or points, as well as changing the line width and fill.

  • Solid Fill:

  • Bars only

  • Points and translucent fill:

Advanced InfluxDB Query Building in Grafana Identifying Metric Series with RegEx

In the example above there were two fairly specific metrics that we wanted to report against. What you will find is much more common is wanting to graph out a set of metrics from the same ‘family’. For example, OBIEE DMS metrics include a great deal of information about each Connection Pool that’s defined. They’re all in a hierarchy that look like this:


Under which you’ve got

Current Connection Count
Current Queued Requests

and so on.

So rather than creating an individual metric query for each of these (similar to how we did for the two session metrics previously) we’ll use InfluxDB’s rather smart regex method for identifying metric series in a query. And because Grafana is awesome, writing the regex isn’t as painful as it could be because the autocomplete validates your expression in realtime. Let’s get started.

First up, let’s work out the root of the metric series that we want. In this case, it’s the orcl connection pool. So in the series box, enter /orcl/. The / delimiters indicate that it is a regex query. As soon as you enter the second / you’ll get the autocomplete showing you the matching series:


If you scroll down the list you’ll notice there’s other metrics in there beside Connection Pool ones, so let’s refine our query a bit


That’s better, but we’ve now got all the Connection Pool metrics, which whilst are fascinating to study (no, really) complicate our view of the data a bit, so let’s pick out just the ones we want. First up we’ll put in the dot that’s going to precede any of the final identifiers for the series (.Capacity, .Current Connection Count, etc). A dot is a special character in regex so we need to escape it \.


And now let’s check we’re on the right lines by specifying just Capacity to match:


Excellent. So we can now add in more permutations, with a bracketed list of options separated with the pipe (regex OR) character:


We can use a wildcard .* for expressions that are not directly after the dot that we specified in the match pattern. For example, let’s add any metric that includes Queued:


But now we’ve a rather long list of matches, so let’s refine the regex to narrow it down:


(Something else I tried before this was regex negative look-behind, but it looks like Go (which InfluxDB is written in) doesn’t support it).

Setting the Alias to $4, and the legend to include values in a table format gives us this:

Now to be honest here, in this specific example, I could have created four separate metric queries in a fraction of the time it took to construct that regex. That doesn’t detract from the usefulness and power of regex though, it simply illustrates the point of using the right tool for the right job, and where there’s a few easily identified and static metrics, a manual selection may be quicker.


By default Grafana will request the mean of a series at the defined grain of time from InfluxDB. The grain of time is calculated automatically based on the time window you’ve got shown in your graph. If you’re collecting data every five seconds, and build a graph to show a week’s worth of data, showing all 120960 data points will end up in a very indistinct line:

So instead Grafana generates an InfluxDB query that rolls the data up to more sensible intervals – in the case of a week’s worth of data, every 10 minutes:

You can see, and override, the time grouping in the metric panel. By default it’s dynamic and you can see the current value in use in lighter text, like this:

You can also set an optional minimal time grouping in the second of the “group by time” box (beneath the first). This is a time grouping under which Grafana will never go, so if you always want to roll up to, say, at least a minute (but higher if the duration of the graph requires it), you’d set that here.

So I’ve said that InfluxDB can roll up the figures – but how does it roll up multiple values into one? By default, it takes the mean of all the values. Depending on what you’re looking at, this can be less that desirable, because you may miss important spikes and troughs in your data. So you can change the aggregate rule, to look at the maximum value, minimum, and so on. Do this by clicking on the aggregation in the metric panel:

This is the same series of data, but shown as 5 second samples rolled up to a minute, using the mean, max, and min aggregate rules:

For a look at how all three series can be better rendered together see the discussion of Series Specific Overrides later in this article.

You can also use aggregate functions with measures that may not be simple point in time values. For example, an incrementing/accumulating measure (such as a counter like “number of requests since launch”) you actually want to graph the rate of change, the delta between each point. To do this, use the derivative function. In this graph you can see the default aggregation (mean, in green) against derivative, in yellow. One is in effect the “actual” value of the measure, the other is the rate of change, which is much more useful to see in a time series.

Note that if you are using derivative you may need to fix the group by time to the grain at which you are storing data. In my example I am storing data every 5 seconds, but if the default time grain on the graph is 1s then it won’t show the derivative data.

See more details about the aggregations available in InfluxDB in the docs here. If you want to use an aggregation (or any query) that isn’t supported in the Grafana interface simply click on the cog icon and select Raw query mode from where you can customise the query to your heart’s content.

Drawing inverse graphs

As mentioned just above, you can customise the query sent to InfluxDB, which means you can do this neat trick to render multiple related series that would otherwise overlap by inverting one of them. In this example I’ve got the network I/O drawn conventionally:

But since metrics like network I/O, disk I/O and so on have a concept of adding and taking, it feels much more natural to see the input as ‘positive’ and output as ‘negative’.

Which certainly for my money is easier to see at a glance whether we’ve got data coming or going, and at what volume. To implement this simply set up your series as usual, and then for the series you want to invert click on the cog icon and select Raw query mode. Then in place of




Series Specific Overrides

The presentation options that you specify for a graph will by default apply to all series shown in the graph. As we saw previously you can change the colour, width, fill etc of a line, or render the graph as bars and/or points instead. This is all good stuff, but presumes that all measures are created equal – that every piece of data on the graph has the same meaning and importance. Often we’ll want to change how display a particular set of data, and we can use Series Specific Overrides in Grafana to do that.

For example in this graph we can see the number of busy connections and the available capacity:

But the actual (Busy Connections) is the piece of data we want to see at a glance, against the context of the available Capacity. So by setting up a Series Specific Override we can change the formatting of each line individually – calling out the actual (thick green) and making the threshold more muted (purple):

To configure a Series Specific Override got to the Display Styles panel and click Add series override rule. Pick the specific series or use a regex to identify it, and then use the + button to add formatting options:

A very useful formatting option is Z-index, which enables you to define the layering on the graph so that a given series is rendered on top (or below) another. To bring something to the very front use a Z-index of 3; for the very back use -3. Series Specific Overrides are also a good way of dynamically assigning multiple Y axes.

Another great use of Series Specific Overrides is to show the min/max range for data as a shaded area behind the main line, thus providing more context for aggregate data. I discussed above how Grafana can get InfluxDB to roll up (aggregate) values across time periods to make graphs more readable when shown for long time frames – and how this can mask data exceptions. If you only show the mean, you miss small spikes and troughs; if you only show the max or min then you over or under count the actual impact of the measure. But, we can have the best of all worlds! The next two graphs show the starting point – showing just the mean (missing the subtleties of a data series) and showing all three versions of a measure (ugly and unusable):

Instead of this, let’s bring out the mean, but still show it in context of the range of the values within the aggregate:

I hope you’d agree that this a much cleaner and clearer way of presenting the data. To do it we need two steps:

  1. Make sure that each metric has an alias. This is used in the label but importantly is also used in the next step to identify each data series. You can skip this bit if you really want and regex the series to match directly in the next step, but setting an alias is much easier

  2. On the Display Styles tab click Add series override rule at the bottom of the page. In the alias or regex box you should see your aliases listed. Select the one which is the maximum series. Then choose the formatting option Fill below to and select the minimum series

    You’ll notice that Grafana automagically adds in a second rule to disable lines for the minimum series, as well as on the existing maximum series rule.

    Optionally, add another rule for your mean series, setting the Z-index to 3 to bring it right to the front.

    All pretty simple really, and a nice result:

Variables in Grafana (a.k.a. Templating)

In lots of metric series there is often going to be groups of measures that are associated with reoccurring instances of a parent. For example, CPU details for multiple servers, or in the OBIEE world connection pool details for multiple connection pools.


Instead of creating a graph for each permutation, or modifying the graph each time you want to see a different instance, you can instead use Templating, which is basically creating a variable that can be incorporated into query definitions.

To create a template you first need to enable it per dashboard, using the cog icon in the top-right of the dashboard:

Then open the Templating option from the menu opened by clicking on the cog on the left side of the screen

Now set up the name of the variable, and specify a full (not partial, as you would in the graph panel) InfluxDB query that will return all the values for the variable – or rather, the list of all series from which you’re going to take the variable name.

Let’s have a look at an example. Within the OBIEE DMS metrics you have details about the thread pools within the BI Server, and there are different thread pool types, and it is that type that I want to store. Here’s a snippet of the series:


Looking down the list, it’s the DB_Gateway and Server values that I want to extract. First up is some regex to return the series with the thread pool name in:


and now build it as part of an InfluxDB query:

list series /.*Oracle_BI_Thread_Pool.*/

You can validate this against InfluxDB directly using the web UI for InfluxDB or curl as described much earlier in this article. Put the query into the Grafana Template definition and hit the green play button. You’ll get a list back of all series returned by the query:

Now we want to extract out the threadpool names and we do this using the regex capture group ( ):


Hit play again and the results from the first query are parsed through the regex and you should have just the values you need:

If the values are likely to change (for example, Connection Pool names will change in OBIEE depending on the RPD) then make sure you select Refresh on load. Click Add and you’re done.

You can also define variables with fixed values, which is good if they’re never going to change, or they are but you’ve not got your head around RegEx. Simply change the Type to Custom and enter comma-separated values.

To use the variable simply reference it prefix with a dollar sign, in the metric definition:

or in the title:

To change the value selected just use the dropdown from the top of the screen:


Another very nice feature of Grafana is Annotations. These are overlays on each graph at a given point in time to provide additional context to the data. How I use it is when analysing test data to be able to see what script I ran when:

There’s two elements to Annotations – setting them up in Grafana, and getting the data into the backend (InfluxDB in this case, but they work with other data sources such as Graphite too).

Storing an Annotation

An annotation is nothing more than some time series data, but typically a string at a given point in time rather than a continually changing value (measure) over time.

To store it just just chuck the data at InfluxDB and it creates the necessary series. In this example I’m using one called events but it could be called foobar for all it matters. You can read more about putting data into InfluxDB here and choose one most suitable to the event that it is you want to record to display as an annotation. I’m running some bash-based testing, so curl fits well here, but if you were using a python program you could use the python InfluxDB client, and so on.

Sending data with curl is easy, and looks like this:

curl -X POST -d '[{"name":"events","columns":["id","action"],"points":[["big load test","start"]]}]' ''

The main bit of interest, other than the obvious server name and credentials, is the JSON payload that we’re sending. Pulling it out and formatting it a bit more nicely:

         "big load test",

So the series (“table”) we’re loading is called events, and we’re going to store an entry for this point in time with two columns, test-id and action storing values big load test and start respectively. Interestingly (and something that’s very powerful) is that InfluxDB’s schema can evolve in a way that no traditional RDBMS could. Never mind that we’re not had to define events before loading it, we could even load it at subsequent time points with more columns if we want to simply by sending them in the data payload.

Coming back to real-world usage, we want to make the load as dynamic as possible, so with a few variables and a bit of bash magic we have something like this that will automatically load to InfluxDB the start and end time of every load test that gets run, along with the name of the script that ran it and the host on which it ran:
SCRIPT=`basename $0`
curl -X POST -d '[{"name":"events","columns":["host","id","action"],"points":[["'"$HOSTNAME"'","'"$SCRIPT"'","start"]]}]' "http://$INFLUXDB_HOST:$INFLUXDB_PORT/db/carbon/series?u=$INFLUXDB_USER&p=$INFLUXDB_PW"

echo 'Load testing bash code goes here. For now let us just go to sleep'
sleep 60

curl -X POST -d '[{"name":"events","columns":["host","id","action"],"points":[["'"$HOSTNAME"'","'"$SCRIPT"'","end"]]}]' "http://$INFLUXDB_HOST:$INFLUXDB_PORT/db/carbon/series?u=$INFLUXDB_USER&p=$INFLUXDB_PW"

Displaying annotations in Grafana

Once we’ve got a series (“table”) in InfluxDB with our events in, pulling them through into Grafana is pretty simple. Let’s first check the data we’ve got, by going to the InfluxDB web UI (http://influxdb:8083) and from the Explore Data » page running a query against the series we’ve loaded:

select * from events

The time value is in epoch milliseconds, and the remaining values are whatever you sent to it.

Now in Grafana enable Annotations for the dashboard (via the cog in the top-right corner)

Once enabled use the cog in the top-left corner to open the menu from which you select the Annotations dialog. Click on the Add tab. Give the event group a name, and then the InfluxDB query that pulls back the relevant data. All you need to do is take the above query that you used to test out the data and append the necessary time predicate where $timeFilter so that only events for the time window currently being shown are returned:

select * from events where $timeFilter

Click Add and then set your time window to include a period when an event was recorded. You should see a nice clear vertical line and a marker on the x-axis that when you hover over it gives you some more information:

You can use the Column Mapping options in the Annotations window to bring in additional information into the tooltip. For example, in my event series I have the id of the test, action (start/end), and the hostname. I can get this overlaid onto the tooltip by mapping the columns thus:

Which then looks like this on the graph tooltips:

N.B. currently (Grafana v1.9.1) when making changes to an Annotation definition you need to refresh the graph views after clicking Update on the annotation definition, otherwise you won’t see the change reflected in the annotations on the graphs.


Everything I’ve written about Grafana so far has revolved around the graphs that it creates, and unsurprisingly because this is the core feature of the tool, the bread and butter. But there are other visualisation options available – “Singlestat”, and “Text”. The latter is pretty obvious and I’m not going to discuss it here, but Singlestat, a.k.a. Sparkline and/or Performance Tiles, is awesome and well worth a look. First, an illustration of what I’m blathering about:

A nice headline figure of the current number of active sessions, along with a sparkline to show the trend of the metric.

To add one of these to your dashboard go to the green row menu icon on the left (it’s mostly hidden and will pop out when you hover over it) and select Add Panel -> singlestat.

On the panel that appears go to the edit screen as shown :

In the Metrics panel specify the series as you would with a graph, but remember you need to pull back just a single series – no point writing a regex to match multiple ones. Here I’m going to show the number of queued requests on a connection pool. Note that because I want to show the latest value I change the aggregation to last:

In the General tab set a title for the panel, as well as the width of it – unlike graphs you typically want these panels to be fairly narrow since the point is to show a figure not lots of detail. You’ll notice that I’ve also defined a Drilldown / detail link so that a user can click on the summary figure and go to another dashboard to see more detail.

The Options tab gives you the option to set font size, prefixes/suffixes, and is also where you set up sparkline and conditional formatting.

Tick the Spark line box to draw a sparkline within the panel – if you’re not seen them before sparklines are great visualisations for showing the trend of a metric without fussing with axes and specific values. Tick the Background mode to use the entire height of the panel for the graph and overlay the summary figure on top.

Now for the bit I think is particularly nice – conditional formatting of the singlestat panel. It’s dead easy and not a new concept but is really great way to let a user see at a real glance if there’s something that needs their attention. In the case of this example here, queueing connections, any queueing is dodgy and more than a few is bad (m’kay). So let’s colour code it:

You can even substitute values for words – maybe the difference between 61 queued sessions and 65 is fairly irrelevant, it’s the fact that there are that magnitude of queued sessions that is more the problem:

Note that the values are absolutes, not ranges. There is an open issue for this so hopefully that will change. The effect is nice though:


Hopefully this article has given you a good idea of what is possible with data stored in InfluxDB and visualised in Grafana, and how to go about doing it.

If you’re interested in OBIEE monitoring you might also be interested in the ELK suite of tools that complements what I have described here well, giving an overall setup like this:

You can read more about its use with OBIEE here, or indeed get in touch with us if you’d like to learn more or have us come and help with your OBIEE monitoring and diagnostics.

Categories: BI & Warehousing