Skip to main content

Notice: this Wiki will be going read only early in 2024 and edits will no longer be possible. Please see: https://gitlab.eclipse.org/eclipsefdn/helpdesk/-/wikis/Wiki-shutdown-plan for the plan.

Jump to: navigation, search

Difference between revisions of "SMILA/FAQ"

(Implementing Pipelets / Processing Services / Bundles)
(Configuring/Running SMILA)
Line 178: Line 178:
  
 
This error usually happens when no index has been build but the user tries to do the search (directly after installing SMILA) by calling the test search application. Prior to 0.5 M3 user gets in this case a long stack trace with the real cause of the problem at the end of it: <tt>org.eclipse.smila.search.index.IndexException: index does not exist in data dictionary [test_index]</tt>. This means that the test index named "test_index" has not been created yet. This can easily be done by starting a crawler or an agent. In 0.5 M3 we improved the error handling in search servlet so that only error messages are displayed and highlighted. (Stack trace can still be seen by simply clicking on "Details...".)
 
This error usually happens when no index has been build but the user tries to do the search (directly after installing SMILA) by calling the test search application. Prior to 0.5 M3 user gets in this case a long stack trace with the real cause of the problem at the end of it: <tt>org.eclipse.smila.search.index.IndexException: index does not exist in data dictionary [test_index]</tt>. This means that the test index named "test_index" has not been created yet. This can easily be done by starting a crawler or an agent. In 0.5 M3 we improved the error handling in search servlet so that only error messages are displayed and highlighted. (Stack trace can still be seen by simply clicking on "Details...".)
 +
 +
==== "Too many open files" error message ====
 +
 +
During indexing process Lucene opens a lots of files which may lead to the error described above. We recommend to increase the number of opened files:
 +
 +
<tt>ulimit –n 16384</tt>
 +
 +
Permanent system configuration changes can be done at:
 +
 +
- System configuration file /etc/security/limits.conf: Add or adapt the limits
 +
for nofile. If it is undefined, the default value is usually 1,024 on most systems.
 +
 +
- System profiles or configuration, e.g. /etc/profile, /etc/initscript, /etc/
 +
profile.local: Check if the ulimit command is used.
 +
 +
- User profiles, e.g. ~/.profile: Check if the ulimit command is used.
  
 
== Implementing Pipelets / OSGi Services / Bundles ==
 
== Implementing Pipelets / OSGi Services / Bundles ==

Revision as of 10:26, 14 June 2011

This pages contains the frequently asked questions of the SMILA project.

General Hint: When you have problems during a SMILA launch / run, please have a look at the SMILA log file first. (<SMILA>/SMILA.log)

Contents

Building SMILA

I receive an Out of Memory error? What can I do?

While building with SMILA.builder I receive the following errror message:

Build Failed - Out of Memory - Java heap space

The reason for this is that Ant hasn´t enough heap space to build the project. You will have to expand the heap space by setting the VM arguments accordingly. In eclipse try the following:

  1. Click Open external tools dialog and select your Ant build profile.
  2. Switch to the JRE tab and add the following VM arguments: -Xms40m -Xmx512m.
  3. Save and build again.


Launching SMILA

Linux

How to start/stop and manage SMILA as a background process on a Linux machine?

Since the default configuration (stored in SMILA.ini) of the OSGi runtime (in our case Equinox) launcher expects that you execute it in foreground and therefore have an OSGi console running in your shell and listening to the standard input, the first thing we have to do is to advise the launcher (and thereby Equinox) to listen on some TCP port instead. This is done by adding a new line with the port number just after the "-console" line.

For example, to set console to listen at TCP port 9999, SMILA.ini would look like this:

-console
9999
...

Now, after SMILA has been started with “$ nohup ./SMILA &”, the console can be accessed from any computer simply by opening a telnet session:

$ telnet <smila_host_name> <console_port>


Symbol comment vote.png If you are logged in via telnet and just want to close this connection and not to stop Equinox running SMILA, than just type “disconnect”. Otherwise, if you want to stop SMILA and close the connection, use “close” - as you normally would do on a console running in your shell.

For the complete documentation on eclipse runtime options please see: http://help.eclipse.org/ganymede/index.jsp?topic=/org.eclipse.platform.doc.isv/reference/misc/runtime-options.html

Bundles

Launching crawler bundles - How to solve: "Could not find crawler id" error message?

While launching SMILA I receive the following error message:

Could not find crawler id

If you started SMILA.launch to launch SMILA: The launcher didn't start your new crawler bundle. Try this:

  1. Add your bundle by selecting "Open Run dialog" in eclipse and choose your SMILA profile.
  2. Select your bundle in the list and set the checkmark.
  3. Set the start level to "4" and the autostart to "true".

If you started SMILA.EXE to launch SMILA: Your bundle isn`t defined in config.ini or the start level isn´t correct. Try this:

  1. Open the file config.ini and add your bundle as shown below:
    org.eclipse.smila.connectivity.framework.crawler.owncrawler@5:start, \
  2. Open the build.properties file of your bundle and include the folders schemas/, OSGI-INF/, and the file plugin.xml.

I changed the implementation of a bundle, deployed it and restarted SMILA, but SMILA still seems to use the old bundle

Close SMILA, delete the following directories in your configuration folder and restart SMILA again:

  • org.eclipse.core.runtime/
  • org.eclipse.equinox.app/
  • org.eclipse.osgi/
  • org.eclipse.update/

I changed bundle settings in my config.ini, but after SMILA restart nothing changed

Check your config.ini for unusual whitespaces (e.g. a tab) between the (edited) bundle entries - and remove them.

If that doesn't help, see question (resp. answer) before.


JConsole

Why is the SMILA package not in the JConsole tree?

I've started SMILA.exe but the SMILA package isn't in the tree of JConsole.

To solve this try the following:

  1. Create a new connection.
  2. Change your connection by setting the port "9004" on the Remote tab.
  3. Click the Connect button, switch to the MBeans tab, and check the tree again.

Why is the SMILA/CrawlerController MBean not in the JConsole tree?

Check if all needed bundles are active. Open the equinox console and type:

  • ss crawler
  • ss deltaindexing.impl

If one of these is not active:

  • check the configuration/config.ini
  • check the log file for errors

Cannot connect with JConsole to a remote machine running Ubuntu 8.04 or newer

For some reason the JConsole cannot make a remote connection to JVM running on Ubuntu 8.04 or newer installations. (We did some tests with SuSE Linux and had no problems.)

This can be easily circumvented by modifying /etc/hosts file. Simply replace 127.0.1.1 with the real IP address of your Ubuntu machine and you're ready to go.

For example, if you have a line in /etc/hosts that looks like this:

127.0.1.1 jupiter (Where "jupiter" it the actual name of the Ubuntu machine in this example. You will almost certainly have some other name here ;-)

then replace it with:

192.168.220.101 jupiter (Where 192.168.220.101 is the actual IP address of the Ubuntu machine in this example.)

Configuring/Running SMILA

Crawler

I tried to crawl/index a data source, JConsole says "Crawl ... sucessfully started" but nothing seems to happen

Check if your queue is receiving records:

  1. Open the JConsole's MBeans tab and go to: org.apache.activemq/localhost/Queue/<queue>/Attributes
  2. if EnqueueCount is increased your queue is receiving records, so it's not a crawler issue.

If your queue didn't receive any records from the crawler:

  1. Did you crawl the data source before? If so, try a clearAll on the DeltaIndexing-MBean in your JConsole
  2. Check your crawler configuration file (see configuration/org.eclipse.smila.connectivity.framework/)
  3. Check the queue routing configuration: org.eclipse.smila.connectivity.queue.worker.jms/QueueWorkerRouterConfig.xml
  4. Check the queue connection: org.eclipse.smila.connectivity.queue.worker.jms/QueueWorkerConnectionConfig.xml
  5. If queue is running in a separate SMILA instance (e.g. on another PC), make sure that "Queue-SMILA" is started before "Crawler-SMILA"

Pipeline

Why do I get a timeout exception during a (long running) pipeline execution?

In SMILA there's a timeout configured for a pipeline execution:

configuration/org.eclipse.smila.processing.bpel/processor.properties

See entry: pipeline.timeout


Lucene Indexing / Search

How can I browse/search an existing Lucene index without SMILA (for debug purposes)?

Try LUKE

Why are attributes in my Lucene index missing / permuted?

Open your configuration files:

  • configuration/org.eclipse.smila.lucene/Mappings.xml
  • configuration/org.eclipse.smila.search.datadictionary/DataDictionary.xml

Check:

  • if all index fields are specified in these two files
  • if the field numbers are compatible to each other

Still the same problem?

  • Close SMILA, remove file (if exists): workspace/org.eclipse.smila.search.datadictionary/DataDictionary.xml and restart SMILA

SMILA doesn't return search results, although I see appropriate entries for (some fields of) my query in the Lucene index

Open the file: configuration/org.eclipse.smila.search.datadictionary/DataDictionary.xml

  • Check the Constraint entries for all fields: If a Constraint is set required and this field is set in your query, the query value must match the indexed value! (regardless of other fields)

Still problems? Try removing the workspace version of that file if it exists:

  • Close SMILA, remove workspace/org.eclipse.smila.search.datadictionary/DataDictionary.xml, and restart SMILA.

Test search application (http://localhost:8080/SMILA/search) returns HTTP Status 500 with the stack trace

This error usually happens when no index has been build but the user tries to do the search (directly after installing SMILA) by calling the test search application. Prior to 0.5 M3 user gets in this case a long stack trace with the real cause of the problem at the end of it: org.eclipse.smila.search.index.IndexException: index does not exist in data dictionary [test_index]. This means that the test index named "test_index" has not been created yet. This can easily be done by starting a crawler or an agent. In 0.5 M3 we improved the error handling in search servlet so that only error messages are displayed and highlighted. (Stack trace can still be seen by simply clicking on "Details...".)

"Too many open files" error message

During indexing process Lucene opens a lots of files which may lead to the error described above. We recommend to increase the number of opened files:

ulimit –n 16384

Permanent system configuration changes can be done at:

- System configuration file /etc/security/limits.conf: Add or adapt the limits for nofile. If it is undefined, the default value is usually 1,024 on most systems.

- System profiles or configuration, e.g. /etc/profile, /etc/initscript, /etc/ profile.local: Check if the ulimit command is used.

- User profiles, e.g. ~/.profile: Check if the ulimit command is used.

Implementing Pipelets / OSGi Services / Bundles

Configuration

I want to use the ConfigUtils class in my Pipelet to read the configuration, where do I have to put my configuration files?

Configuration files are searched for in the following order:

  1. <SMILA>/configuration/<bundle-name>/<config-file>
  2. <config-file> in the root path of the bundle jar-file

Deploy / Launch

I implemented/deployed a OSGi Service in a new bundle but SMILA log says that it couldn't be found

Check your new bundle, it should contain a file like that:

  • OSGI-INF/<myService>.xml

In this file your new service has to be referenced. If you have copied the file from some other service, be sure to change the component name in the root element to something unique, because DS does not start multiple services with the same component name.

  • <component name="<myService>" immediate="true">

Also the file has to be referenced from the MANIFEST.MF file of your bundle as a service component:

  • Service-Component: OSGI-INF/<myService>.xml

Also, you may need to include Import-Package: declarations for super-classes of your service implementation class even if there are no compile errors.

On the "Build" page of the manifest editor, you must add the OSGI-INF directory to the binary build.

And finally, your bundle has to be started at SMILA launch, e.g. by adding it to the config.ini.

I get classloading errors in invocations of my own Pipelet when running SMILA outside the IDE. In the IDE it works

The error could look like this:

2010-11-19 11:28:36,101 ERROR [ODEServerImpl-1 ] vpu.JacobVPU - Method "run" in class "org.apache.ode.bpel.rtrep.v2.EXTENSIONACTIVITY" threw an unexpected exception. java.lang.LinkageError: loader constraint violation: loader (instance of org/eclipse/osgi/internal/baseadaptor/DefaultClassLoader) previously initiated loading for a different type with name "org/w3c/dom/Document"

We are not completely sure, why this happens, but a solution is to set this system property in the SMILA.ini file:

-Dosgi.java.profile.bootdelegation=override

Thamks to Bogdan Sacaleanu for the solution. See this thread in the smila-dev mailing list for additional details.

Back to the top