[
https://issues.apache.org/jira/browse/ROL-2094?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15044119#comment-15044119
]
David Johnson edited comment on ROL-2094 at 12/13/15 2:21 PM:
--------------------------------------------------------------
One way to make the Lucene index distributed is to use Apache Solr.
Solr is a distributed search engine that is built on top of Lucene. We would
have to rewrite the portions of our code that work with Lucene and change them
to call the Solr API. Since Solr is based on Lucene and Lucene concepts, this
should be straight-forward.
Users who want to setup Roller to run in a distributed environment would have
to setup a Solr cluster and users who want to use Roller on one machine would
configured Roller to use Solr embedded.
was (Author: djohnson):
One way to do this is to use Apache Solr.
Solr is a distributed search engine that is built on top of Lucene. We would
have to rewrite the portions of our code that work with Lucene and change them
to call the Solr API. Since Solr is based on Lucene and Lucene concepts, this
should be fairly simple.
Users who want to setup Roller to run in a distributed environment would have
to setup a Solr cluster and users who want to use Roller on one machine would
configured Roller to use Solr embedded.
> Making Roller capable of clustered environment
> ----------------------------------------------
>
> Key: ROL-2094
> URL: https://issues.apache.org/jira/browse/ROL-2094
> Project: Apache Roller
> Issue Type: Improvement
> Components: Installation & Configuration
> Affects Versions: 5.1.2
> Reporter: Kohei Nozaki
> Assignee: Roller Unassigned
> Priority: Minor
>
> If my understanding is correct, Current version of Roller doesn't work well
> in clustered environment as is. I think it would be great if Roller is
> capable of large-scale, distributed and heavy-workload usecase, With
> deployment architecture such as described in
> http://image.slidesharecdn.com/whats-new-in-roller5-091124062729-phpapp01/95/whats-new-in-roller5-7-728.jpg?cb=1259044066
> .
> Original posting:
> http://mail-archives.apache.org/mod_mbox/roller-user/201505.mbox/%3cCAPyOjHhrOwhX=BhV+Wnv6N2M_55SDXFF6xHapiOc=7_xmvj...@mail.gmail.com%3e
> As far I know the hurdles for clustered environment is following:
> # Local media directory
> # Lucene Index
> # Page Cache
> # Salt processing
> # 2nd Level cache of EclipseLink
> Workarounds or things we need to get done for each hurdles above:
> # Using NFS would help?
> # No idea, But I guess it's a common requirement so well established
> solutions should exist
> # Implement cluster-aware one of
> {{org.apache.roller.weblogger.util.cache.CacheFactory}} ? For example, Based
> on Memcached?
> # Implement 3 and make salt processing aware of it?
> # Just put {{<shared-cache-mode>NONE</shared-cache-mode>}} in
> {{persistence.xml}} would help, But it degrades performance and most of
> installations will work as standalone, So adding a note to documentation
> (e.g. installation guide) for clustered environment would be reasonable
> I'm still not aware of problems or solutions to it enough but this is just a
> starting point. Any advices or opinions would be appreciated.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
