Upgrading Solr

[anghelutar]

Hello,

I am working on a project to upgrade the Solr version used by Alfresco. Current version is 1.4, which is really old, latest being 4.3. There are many new features, amongst which the possibility to distribute the index, more advanced faceting and performance improvements.
Is there anyone who already did this or is interested in sharing the work? Is Alfresco interested in shipping out Alfresco with a newer version of Solr?

roxana

[jpotts]

Yes, there are many people interested in seeing this happen and in doing the work. It is a very large effort with several dependencies involved. It is on our roadmap and we are actively talking about how to get this done.

[anghelutar]

Would it be possible to be involved in this process, is there a Jira ticket for it? I actually already started to work on the upgrade and I wouldn't want to end on a dead-path because the official version takes another turn.

Thanks,
roxana

[jpotts]

I've given a heads-up to the internal Engineers and the external partner that have been planning this work. Hopefully they will weigh in here and see if there is a possibility to get you involved and perhaps leverage what you've done so far.

Big changes like this definitely need to be coordinated with engineering or they run the risk of being obsoleted, just as you point out.

Jeff

[matteo_grolla_2]

Hi Jeff and Roxana,
   I'm Matteo Grolla from Sourcesense, an Alfresco partner in Italy and UK. I've done some analysis on the current architecture of the connector and have started discussing with Alfresco engineers the necessary changes to support a modern version of Solr.
I'm very interested in sharing thoughts about the connector upgrade.

[anghelutar]

Hi Matteo,

My starting point was to try to compile the code from svn against the new libraries, without changing the architecture really. The modules affected are: data-model, repository and solr.
Besides smaller things, up to now I've hit problems in 3 main sections:
1) different ways of handling TokenStreams (see for example http://stackoverflow.com/questions/2638200/how-to-get-a-token-from-a-lucene-tokenstream)
2) different ways of handling Enums (see http://lucene.apache.org/core/4_0_0-BETA/MIGRATE.html)
3) snowball analyzers are not supported anymore
With a bit of work, I managed to compile the classes I looked at and I am confident that this work can be actually done for all existing classes. But I stopped because I realized that I might diverge from Alfresco, it would be best to first see whether somebody else is looking at this and join efforts.

roxana

[matteo_grolla_2]

I've sent you my skypeid as a private message so we can have a talk if you want.