Formerly named OODT Easy Install
ApacheCon NA 2011 Fast Feather Talk
Info from the ApacheCon NA 2011 talk by Paul Ramirez and Cameron Goodale is here: http://lanyrd.com/2011/apachecon-north-america/smfcf/
The high level goal of this effort is to build a distribution of OODT that sets up, installs, and runs within a few commands. Reducing the number of commands to run OODT is an effort to push the edge of ease of setup and configuration required to get going. This distribution of OODT will include both a deployment and source structure for managing the evolution of your installation of OODT.
Maven offers an improved way to export, configure and build OODT called Archetypes. Archetypes simply put are a way to define templates for projects. Within these project templates we will include packaging instructions to conform to the guidelines below to increase the similarity amongst deployments of OODT. Moreover, we will build higher level scripts and configuration to tie the pieces together at the system level. Finally, we will leverage the CAS Install Maven Plugin to take us from our source structure to our deployment structure.
This wiki will be used to capture thoughts, ideas and plans for the first archetypes we develop for OODT. To keep things simple we are going to initially focus on a small number of modules that are typically deployed and configured together. Finally, our goal is to build an 80% solution that works in most cases to get people out of the gates and running with a full OODT solution. We believe this effort will help increase adoption and conformity amongst installations of an already great system.
Long, long, ago there was a thought that making some archetypes for OODT would be great to get people going. Of course we already had example projects along with each of our components but let's be honest, how many of us started from those? Okay, so a few did but that's not the point. Sometime later a wise man pondered that it might be cool to borrow the idea Cloudera like distribution for OODT. That of course fizzled, until sometime recently a young padawan came along and said OODT should just be easier. The archetype guy said, you are right and that's what I've been saying. Of course, being that the idea was always in the aether and not a reality no one knew. That is until now. RADiX is the realization of what has been brewing and itch that has been waiting to be scratched. It builds on stuff we have done all along but brings our work to the forefront and helps it shine. This new way of working with OODT will not be for everyone but we hope that it will give the 80% solution and that,"Then you'll see, that it is not the spoon that bends, it is only yourself."
- The initial archetype will export RELEASED versions of OODT
- The initial archetype will export Crawler, FileManager and Workflow Manager ONLY (they will be bundled together and configured to work together)
- other modules will be added in the future
- FileManager Policy will be read recursively from the components/filemanager/policy directory. This will remove the requirement to make properties updates when additional policy files are added in sub-directories.
- Archetype will only support a single version for all components. No mixing versions of individual components
- Maven Archetype process will be completed using as few commands as possible
- Submit INFRA ticket to create a place to put all Maven Central artifacts
- Load/Install artifacts to Maven Central
Requires Maven 2.x and Java 1.5+
The archetype is available at maven central. Here is the set of commands you would typically run.
You should then edit the radix file replacing the parameters below with whatever you want.
The first command is the running of a maven archetype to make an oodt project. Inside of the RADIX script downloaded in the curl command is an mvn archetype generation command. That command has a number of parameters that the 1 liner radix script encapsulates. (parameters marked in italics below)
The groupId is a place to specify your company's namespace.
The artifactId is a place to specify a short name of your project.
The version indicates the initial version label for your project.
The oodt flag indicates the version of OODT that you want your project to be built on. N.B., this should most likely match the most recent version of OODT.
- The second command simply moves into the created oodt project directory where your source and configuration can be maintained and later can be placed into version control. Then creates the distribution of your OODT project using "mvn package"
- The third command merely untars the distribution into the created deployment directory
- The fourth command moves into the deployment directory and starts the OODT system.
- The fifth command launches batch stub on the port 2001.
NOTE1: If you observe for some reason OODT doesn't start, make sure JAVA_HOME is set in your ~/.bashrc (Example: export JAVA_HOME="/usr/lib/jvm/java-6-openjdk-i386") and then start OODT in a new terminal. The actual problem can be seen in the tomcat logs which is located in $OODT_HOME/tomcat/logs.
NOTE2: After you launch oodt (the fourth command), you may observe the following output:
Don't be confused. In order to see whether the oodt is running, open a browser to http://localhost:8080/opsui. Click on PCS Status link to get detailed information about running processes. A green arrow indicates that the corresponding process runs correctly.
Alternatively one may run the following command and get list of relevant processes and assigned ports.
The "oodt start" command should return quite quickly. This is normal. If, however, the http://localhost:8080/opsui link above is not functional on your system, one good place to start looking for probable cause is the "oodt.out" log file, which can be found (if you've been following the example above) in "oodt/logs/oodt.out"
OODT Start and Stop Issues
While starting/stopping oodt if you get errors like "Is File/Workflow/Resource Manager
still running? Start/Stop aborted." in logs/oodt.out:
Remove any *.pid files (example:cas.filemgr.pid) from the run directory of
File/Workflow/Resource Manager. This will make sure that the corresponding
process has been stopped.
To check if all managers are running as expected, we can use the http://localhost:8080/opsui as stated above
or simple try using these links in browser
You should get this message for all these three links: "Method GET not implemented (try POST)" which means all
those managers are running. Note that you can also check if the managers are running as expected by visiting
$OODT_HOME/pcs/scripts/ and by running
If you want to manage your OODT RADiX distribution with Subversion.
If you want to manage your OODT RADiX distribution with Git
prompt> git init
prompt> git add .
prompt> git commit -a -m "Initial OODT import"
Default Deployment Structure
In order for the easy installation to work properly we will need to settle on a default deployment layout structure. Below is our plan of how we want to layout the deployment when the project is built. First we will list out an overview, then we can detail each path and what files are going to be saved into each.
Deployment Path Descriptions
This is the root of where the filemgr will store its archived products
This directory will be monitored by the crawler. Products to be ingested should be placed here
In a configuration that uses Lucene as a back end this directory holds the contents of that index
Any products that have failed ingestion will be placed here along with any metadata files.
System level scripts to start, stop, restart the OODT infrastructure.
Crawler deployment for your data management system (i.e. policy, scripts, and configuration). This component is responsible for monitoring the staging area
Filemgr deployment for your data management system (i.e. policy, scripts, and configuration). This component catalogs and archives products into the archive area.
Workflow deployment for your data management system (i.e. policy, scripts, and configuration). This component orchestrates any processing that may need to be done on your products
Sandbox area to test out metadata extractors, versioners, actions, etc. that you have developed to extend the functionality of the existing OODT framework.
System wide configuration
This directory contains scripts manipulate the underlying components. You can start, stop, and restart each individual component or all 3 at the same time.
Manipulate all components (DEFAULT BEHAVIOR):
Manipulate a single component:
These are the default parameters set within oodt.properties.
Component settings we plan to default
Default Source Structure
Source Path Descriptions
Project specific crawler configuration, policy, and scripts
Project specific filemgr configuration, policy, and scripts
Project specific workflow configuration, policy, and scripts
Web Applications from Apache OODT
Extensions to the OODT framework to do metadata extraction, archive layout (aka versioner), workflow tasks, crawler actions
Distribution package project for system level build, configuration, and scripts
Once the above is complete our thoughts are that the next items to be incorporated are as follows:
- Tomcat Distribution
- OODT Services (Health Monitor, ?)
- OODT Web Apps (Curator, ?)
- CAS PGE
- Expand OODT Easy Commands
- upgrade - to allow for upgrades in OODT components
- status - to print out the version of OODT running and component status
- add_product_type - to configure all components with a new product type
Maven Archetype Information
Requirements to getting Artifacts Synched with Maven Central: