OBIEE 11g installation – Part3
I ran the installer, and selected ‘simple install’. The prerequisite check came up with a warning that my physical memory was insufficient to run the application. I’m going to ignore this for now, since I don’t want to allocate any more memory to this virtual machine given that I need to run a db simultaneously.
I selected c:\obiee as the Middleware home – this would be the install dir for OBIEE 11g. I got an error message right away that the MW_HOME required 11,264 MB of free space and I had far less. This is definitely a departure from OBIEE 10 where the entire app (minus cache and temp files) could easily fit in under 3GB.
I cancelled the installation, shutdown the virtual machine, and added another 20gb drive. The next stage of the install asked for a system admin password. The default userid was Weblogic. I changed this to Administrator.
On the component selection screen, I selected both Business Intelligence and Publisher. I’ll hold off on Real-Time Decisions for a later blog post.
On the Database Details screen I entered the same credentials as during the rcu phase. The connect string needed to be entered as hostname: port:instance name. The userid for BIPLATFORM was prefix underscore BIPLATFORM: DEV_BIPLATFORM. It took a few tries to get this right (I ended up checking enterprise manager on my db for the userid as I had not written it down from the previous step). Ironically entering in sys for the userid yielded an error message that the user did not have sysdba rights… Also, its important to note that there are only 3 supported database types for the repository, and those are Oracle, SQL Server, and DB2.
I did not specify an email on the Specify Security Updates screen. I also unchecked ‘receive updates’ as this is a demo build. You can actually move forward without filling in these details, provided you accept the ‘Do you wish to remain uninformed…’ disclaimer.
The next screen appears to be the final screen in the configuration process for ‘simple install’. I went ahead and saved the ‘file’ prior to continuing. I then hit next and here we are – the installation is off and running.
Now, I have not tried consolidating all of the uncompressed files into a single folder structure, but the installer seems to be picking up the required files from their individual uncompressed folder locations. I stand corrected. Midway through the install process, the installer prompted for the location of Disk 2. At this point, I consolidated all the uncompressed files into the first (and current install folder). This worked, and the installer completed step 10 without further incident.
And there we have it.
Lessons learned?
Start with a database, and the RCU utility. There is no need to create any schemas up front, use the sys account to generate the other repository schemas using rcu.
Consolidate all of the compressed installation files into one folder structure prior to starting the install.
Make certain you have enough space for the ~5GB of install files, and ~11GB required space for the actual installation.
The ‘Simple Installation’ by default installs Weblogic which is now an integral part of OBIEE 11g. All configuration management is done via Enterprise Manager, which runs off of WebLogic.
After the software installs, the installer goes through a configuration cycle:
This is an automated process as part of this install type.
It’s been a good 5 minutes, and the configuration is sitting at ‘Restarting Admin Server’. One hour later, the install had not progressed. There had been no updates to the log files, or any movement whatsoever within the install folders.
I cancelled the install, uninstalled Weblogic, restarted the virtual machine (first having added another ½ GB RAM) renamed the install folder, and restarted the installer. Even though the installation folder had been renamed, I received an error in the last stage that an Oracle home already existed, and that I needed to select a different home. Unfortunately there was no way to ‘go back’ in the install, so yet again I had to cancel the process. This time, I deleted the original installation folder instead of just renaming.
Upping the virtual machine’s RAM to 1.5 GB caused the prerequisites to pass on the memory check. I kept all the install settings/parameters as before. The installation went smoother this time, failing near the end at the ‘Executing: opmnctl start coreapplication_obiccs1’ task. On doing a bit of searching through the logs, this task appears to be related to the cluster controller – which does not apply to this install. Although not intuitive, the continue button ignores the failed step and continues the configuration phase.
Unfortunately, the opmnctl start… ‘OracleBIPresentationServicesComponent’ task failed as well. This time is was due to ‘Unable to connect to port 9706 on machine…’. Launching the Weblogic service did not resolve the issue. Reviewing the log files contained in the ‘E:\obiee\instances\instance1\diagnostics\logs\’ folder it appears that a number of configuration options have not been properly set by default – for example, several commands reference a path to ‘\\nullserver\’ as part of the cluster configuration.
I selected continue to skip over the second of the failed tasks, and the other configuration tasks completed without fail. In summary, the presentation and cluster servers failed to launch during the configuration process.
And there we have it, a completed, albeit not without issues, installation of OBIEE 11g using the ‘Simple Installation’ option.
I can tell you after the fact that a number of configuration steps could have been done differently in the ‘simple installer’. For example, the default installer attempted to create an application cluster (including configuring the default system DSN to be clustered) which might be better left to the advanced/normal installation. However I am not sure whether there is a non-clustered install options, and as an afterthought it may be that Oracle has done this for consistency i.e. a single machine install is just a one node cluster.
I can also tell you that the rpd file which we all know and, well, we all know, is still there, as is nqsconfig.ini, instanceconfig.xml, etc… Most files have been shifted around, but the application retains many of the core components we are familiar with – NOTE: YOU CANNOT MODIFY THE CONFIG FILES MANUALLY. All configuration and setting changes must be performed through Enterprise Manager, if you try to modify the config files manually, your changes will be overwritten the next time you start BI.
I do like how Oracle has set separate folder structures for ALL config files, log files/diagnostics, and repository/catalog files. There is much more consistency to the 11g which makes the application more logical to follow. A fact I’m grateful for since the next step for me is to troubleshoot the failed installation steps in order to bring up the front end.
For example, while troubleshooting I did a stop of BI Applications, and a start of BI Applications (ensuring WebLogic was up and running), altered the system DSN to non-clustered, modified \instances\instance1\config\OracleBIApplication\coreapplication\nqclusterconfig.ini and disabled clustering, and changed the instanceconfig.xml file (the original had a mix of / and \ slashes in the catalog path, I modified the slashes to be consistent with a Windows environment. All of these changes were lost when I did another restart of BI Applications.
The OBIEE 11g front end now comes up via the standard URL
http://localhost:9704/analytics. It took a little while for the front end to load on my laptop, and I did initially get an HTTP error while loading the page. This is likely due to the fact that my laptop is underpowered to run all of these components in a virtualized environment (I have run this entire process on my Dell E6400 laptop (dual core, 4GB RAM)). However, with a bit of patience, all components eventually started and I did get to the login page.
I received an error logging in that my userid and password could not be authorized. This was a standard error message and I thought that it indicated that I did not have the correct password at first. On doing a bit of research, its apparent that Oracle now uses WebLogic’s identity management tools, and that the primary weblogic account now acts in place of the Administrator account when connecting to the rpd or via the front-end. This is not to be confused with a new security feature called the repository password.
A bit of troubleshooting later I noticed that I could not connect to the rpd in online mode. Instead of getting the unable to authenticate error, this time I received a message that the cluster controller could not be reached. This led me to believe that the failed install step (cluster configuration) may be causing this issue.
I thought that found the issue… The installer calls a file called D:\obiee11g\Oracle_BI1\opmn\bin\opmnctl.bat which incorrectly had concatenated the PERL5LIB assignment with the opmnctl command. I was wrong. I traced through the batch files and found the Perl script they called. Analyzing the Perl script I found that all the batch files and Perl script did was to call opmn.exe like so: opmn.exe -a start ias-component=coreapplication_obiccs1.
The call to opmn starts the cluster controller. I changed the Perl script to run opmn.exe in verbose mode and received the following error message:
D:\obiee11g\Oracle_BI1\opmn\bin>D:\obiee11g\Oracle_BI1\opmn\bin\opmn.exe -av sta
rt ias-component=coreapplication_obiccs1
HTTP/1.1 204 No Content
Content-Length: 695
Content-Type: text/html
Response: 0 of 1 processes started.
<?xml version=’1.0′ encoding=’WINDOWS-1252′?>
<response>
<opmn http-status=”204″ http-response=”0 of 1 processes
started.”>
<ias-instance>
<ias-component>
<process-type>
<process-set>
<process pid=”3216″ status=”Init” index=”1″ log=”" oper
ation=”request” result=”failure”>
<msg code=”-21″ text=”failed to start a managed process after the ma
ximum retry limit”>
</msg>
</process>
</process-set>
</process-type>
</ias-component>
</ias-instance>
</opmn>
</response>
I have been doing a bit more troubleshooting to get the cluster controller to start – among other things I made sure that the repository was published to a shared location (it was not originally, and the documentation did say that for clustered deployments it should be).
Although still troubleshooting, I really wanted to see the new front end, so I tricked the system (in a manner of speaking). We know that the 11g overwrites standard parameters and configuration settings such as ODBC settings when the BI services start (using the start BI services batch file). However, I was able to change the OBIEE ODBC to non-clustered mode, and simply restart the presentation and bi server services through EM without doing a full component shutdown and restart. This allowed me to log into the front end of the SampleAppLite application.
Oracle says that the SampleAppLite is a small scale version of SampleApp, and the full SampleApp is available for download from OTN – I was not able to find it when I searched earlier, and then things got too busy throughout the day and I did not have a chance to go back to find it… However, just looking at SampleAppLite I have to say that I was pretty impressed.
Ok, more than impressed. I had a big grin on my face. Everything from the Flash-enabled chart sliders, to dynamic pivot charts (yes, you can drag your columns around) was a fantastic improvement to OBIEE 10.
I’m going to get back to working on the cluster controller issue and will update this posting with what I find.
It’s been a few hours, and I have a solution – I set up a loopback adapter with a static IP and set it as the first network interface (in order of adapters). The lesson to be learned is that OBIEE 11g seems to require a static IP – similar to Oracle database. I’ll update the first blog post in this series to point this out.
As for the next blog, its largely been written, I was just waiting to resolve the above mentioned issue. As a preview: the blog is about creating a demo environment – using a smaller footprint. Here’s an exerpt:
“ OBIEE 11g running on my virtual machine (on my laptop) with SQL Server 2008 Express hosting the rcu repository actually runs quite well. The services take a while to start up, but once they are up and running, report performance is pretty good, even more complex elements like the scorecard come up in a reasonable amount of time. “
