Paul Piper created OFBIZ-5042:
---------------------------------
Summary: Apache Solr Implementation
Key: OFBIZ-5042
URL: https://issues.apache.org/jira/browse/OFBIZ-5042
Project: OFBiz
Issue Type: Improvement
Components: ALL COMPONENTS
Reporter: Paul Piper
Apache Solr, a modern search engine, seems to be one of the most commonly
requested feature within the OFBiz community. With my agency, ilscipio, we have
done several Apache Solr implementations in the past, and we felt that we could
contribute this feature to the community. Currently, Solr is greatly used
within syracus.net (our standalone store) and serves to generate the categories
& search functionalities. From our team mainly Pascal Proulx, Jonatan Soto and
I were involved in this feature (albeit more having contributed to it. We
created a standalone component which can be integrated into any running ofbiz
implementation. It was tested against OFBiz 09.04 - 12.04..
=====================================================
Solr Component for Apache Ofbiz
=====================================================
This document describes the Ofbiz solr component, an Ofbiz
(http://ofbiz.apache.org/)
implementation of the Apache Solr search platform
(http://lucene.apache.org/solr/).
The solr component includes an Ofbiz service-based wrapper layer to the Apache
Solr
webapp queries as well as the native Apache Solr web interface itself.
Currently, the solr component focuses on Product data.
Note: This document is a work in progress; information is subject to change.
Contents:
1. Installation
2. Configuration
3. Data Indexing
4. Data Querying
5. Implementation Concerns
6. Known Bugs, Limitations and Issues
-----------------------------------------------------
1. Installation
-----------------------------------------------------
To install solr in an Ofbiz setting, simply extract the solr directory and
and sub-folders to the hot-deploy folder.
Afterward, the solr home system property (solr.solr.home) must be set to the
value
hot-deploy/solr manually using one of the following methods:
* solr.solr.home in batch/script file:
Add the parameter "-Dsolr.solr.home=hot-deploy/solr" to the Java command
invocation for ofbiz.jar.
e.g.:
"%JAVA_HOME%\bin\java" -Xms128M -Xmx512M -XX:MaxPermSize=512m
-Dsolr.solr.home=hot-deploy/solr -jar ofbiz.jar
* solr.solr.home in Ant build configuration:
In your root Ofbiz build.xml file, add the element "<jvmarg
value="-Dsolr.solr.home=hot-deploy/solr"/>"
to the "<java jar="ofbiz.jar"...>" invocation of the appropriate Ant target(s)
(run, start, run-install, etc.).
e.g.:
<target name="start" description="Start OFBiz">
<java jar="ofbiz.jar" fork="true">
<jvmarg value="${memory.initial.param}"/>
<jvmarg value="${memory.max.param}"/>
<jvmarg value="${memory.maxpermsize.param}"/>
<jvmarg value="-Dsolr.solr.home=hot-deploy/solr"/>
</java>
</target>
***
It may be possible to specify solr home using other methods (JNDI, web.xml),
but at the time
of this writing, this was the most reliable method known.
-----------------------------------------------------
2. Configuration
-----------------------------------------------------
The solr component can run out-of-the-box without configuration, but many
files, settings and interfaces allow custom settings. Some of these include:
* Ofbiz configurations:
** System properties:
*** ofbiz.solr.eca.enabled - Global solr ECA toggling boolean (true/false, see
Data Indexing)
** Config files:
*** solr/config/solrconfig.properties - Ofbiz solr service behavior control
*** ofbiz-component.xml - Standard Ofbiz component config
* Apache Solr configurations:
** System properties:
*** solr.solr.home - Solr home (see Installation)
** Config files:
*** solr.xml - Base solr config
*** conf/schema.xml - Solr index schema
*** webapp/WEB-INF/web.xml - Dual Ofbiz/Solr webapp config
** Interfaces:
*** /solr/admin/ - Webapp admin interface (see below)
***
It is possible to set extensive native Solr configuration using
the admin webapp interface noted above. It should be accessible at the address:
http://localhost:8080/solr/admin/
(where 8080 is your server's http port)
Please refer to the Apache Solr documentation for usage of this interface
and other native Solr configuration details.
-----------------------------------------------------
3. Data Indexing
-----------------------------------------------------
The solr component indexes data such as Products into the Apache Solr database
using services defined in the file:
servicesdef/solrservices.xml
The initial indexing may need to be performed or scheduled manually, but
subsequent indexing
may be partially or fully automated, though automated methods are disabled by
default and must be enabled.
Note that in general, solr services can only successfully run in contexts where
the solr webapp is
loaded and accessible.
There are two methods for indexing data:
* Index rebuilding service (rebuildSolrIndex):
The rebuildSolrIndex is the most important data import service. It reindexes
all Ofbiz Products existing in the system into the solr index. rebuildSolrIndex
MUST be run
AT LEAST once after installation and also following any data load operation
that loads new products using the Ofbiz
"install" starting mode (run-install, load-demo, etc.).
To do this, one can simply use the Webtools backend to invoke the service
manually
(e.g.,
/webtools/control/setSyncServiceParameters?SERVICE_NAME=rebuildSolrIndex&POOL_NAME=pool&_RUN_SYNC_=Y)
or summon the Webtools Job Scheduler.
Once the initial indexing has been performed, one can then use the Job
Scheduler to
invoke rebuildSolrIndex on a regular basis (every hour, every midnight, etc.)
to update the Solr index.
* ECAs/SECAs (addToSolr, for Product data):
Although the rebuildSolrIndex is always necessary for the initial data import,
one may also
use ECAs and SECAs to import subsequent data changes automatically at every
individual data (e.g. Product)
update instead of running rebuildSolrIndex periodically. This is done by
defining ECAs or SECAs that
trigger the addToSolr service.
The addToSolr service simply accepts a single "instance" parameter, a
GenericValue. At the time of this writing,
any entity value having a valid "productId" field designating a Product value
may be passed; this will trigger reindexing
for the specific product.
By default, the addToSolr service implementation (and any service intended for
use with ECAs) is disabled globally and
succeeds silently. It can be enabled by setting the "solr.eca.enabled=true" in
"config/solrconfig.properties".
This property is provided for easy toggling. It can also be specified on the
command line or using ant through
the true/false "ofbiz.solr.eca.enabled" system property (which has priority
over file-based "solr.eca.enabled") in the same
way described for the solr home system property (see Installation); this allows
toggling them per Ant target.
Some stock/default/example ECAs are provided and are functional once enabled.
They can be found in:
entitydef/eecas.xml
Custom ECAs may also be added to this file, and custom SECAs may be added to
the file:
servicedef/secas.xml
Special care may need to be taken to ensure the addToSolr is not triggered
during "install" sequences
(run-install/load-demo) because it requires the solr webapp to be accessible.
addToSolr will automatically
attempt to prevent reindexing in cases where the solr webapp is unavailable.
However, it is possible that
this behavior results in needlessly prevented reindexing (and thus lost
updates) in some unknown complex server
configurations. If this is the case, this behavior can be disabled by setting
the option
"solr.eca.useSolrWebappLoadedCheck=false" in the properties file
"config/solrconfig.properties".
However, if the above useSolrWebappLoadedCheck option is specified, you will
have to manually disable or comment the
ECA definitions/services to prevent them from executing during data-load
operations; the "solr.eca.enabled" property and "ofbiz.solr.eca.enabled" system
property can be used for this purpose. One easy way is to set
"solr.eca.enabled=true" in
the property files and then specify "-Dofbiz.solr.eca.enabled=false" for any
relevant
Ant build install/run-install/load-demo targets.
Note that simply commenting the ECAs/SECAs or preventing their inclusion using
ofbiz-component.xml may eliminate some service call overhead compared to using
the property toggling. Ideally,
a method exploiting ofbiz-component.xml or eecas.xml to prevent ECAs from
running in inappropriate contexts
would be ideal; however, at the time of this writing, such a method could not
be established.
***
To reset and clear the solr index completely, it is sufficient to shut down the
server and delete the folders:
solr/data/index
solr/data/spellchecker
and any other such folders (must delete the whole folders, not only contents).
-----------------------------------------------------
4. Data Querying
-----------------------------------------------------
Solr queries can be done using two methods:
* Solr Ofbiz services:
Simply invoke (manually or in code) the query services found in the file:
servicesdef/solrservices.xml
These include solrProductsSearch, solrKeywordSearch and others. Note that in
general,
solr services can only successfully run in contexts where the solr webapp is
loaded and accessible.
* Solr native admin webapp interface:
One can also perform native Solr queries and diagnostics using the standard
admin interface, accessible as described under Configuration. Please refer
to the Apache Solr documentation for usage of this interface.
-----------------------------------------------------
5. Implementation Concerns
-----------------------------------------------------
* The structure of the solr/webapp directory closely mirrors the contents of
the solr.war
distribution; however, please note that it contains some Ofbiz-specific
modifications such
as those found in web.xml. The same is true for other folders such as
solr/conf. Therefore,
any library updates must be done with care to preserve these modifications.
-----------------------------------------------------
6. Known Bugs, Limitations and Issues
-----------------------------------------------------
* In general, solr services can only successfully run in contexts where the
solr webapp
is loaded and accessible.
* ECA indexing may be problematic due to data loading and unknown complex
server configurations;
various toggling options are provided to work around this issue (see Data
Indexing for details).
In the future - and ideally - this could be addressed using ofbiz-component.xml
(or eecas.xml) directives.
* The indexing services often produce the log warning "Problem reading product
features" due to their
use of the stock Ofbiz getProductFeatureSet service.
***
Please report any other issues encountered. Thank you.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
