Migration fom Lucene to Solr in a high load production system

[prim]

We are trying to migrate our Alfresco Enterprise 4.1.6 system from Lucene to Solr indexing.

I've created a clone of the production system that uses the same hardware and made a full-index (Lucene) to ensure that there are no hidden problems, it has taken 84 hours with no problems and everything seems working OK.

I followed the documentation Installing and Configuring Solr performing all steps smoothly, in a separte Solr server, and the system has started solr indexing.

I have serious doubts that we can migrate the real production system:

- The summary solr query indicates that it will take more than 20 days

- While solr indexing, alfresco repository server is running but the time documenbts loads take too long, sometimes resulting time-out.

- This will prevent productive run this migration, we are adding more than 9,000 objects daily and these response times will be impossible.

- I need to unstress the repository server enough to permit the daily documents load, even if the solr index takes 30 or 40 days.

What can we do to migrate?

Can we run this process impacting less on performance and make it compatible with the daily workload?

Is there any alternative provided for more or less large environments like ours?

I need your help or suggestions of experts to address this migration.

Alfresco Enterprise 4.1.6

MySql server: 2Cpu, 4GB ram, 225Gb disk.
Tomcat server: 4Cpu, 11Gb ram, 940Gb disk.
Solr server: 4Cpu, 11Gb ram, 130Gb disk

Best regards<response>

[prim]

http://xx.xx.xx.xx:8080/solr/admin/cores?action=SUMMARY&wt=xml

[mrogers]

Please contact alfresco support.

[prim]

I've done it, I've an open case (00319019) since 12 feb, but have no answer, only a rapid and banal response (so the ticket has "a good time response"), Alfresco support doesn't seem be the best support.

Is this forum only for community users, not for Enterprise?, excuse me.

Thank you

[openpj]

<blockquote>Is this forum only for community users, not for Enterprise?, excuse me.</blockquote>
Typically Enterprise users should use the official Alfresco Support for similar problems.
Anyway questions about the Enterprise version are also welcome here

I think that probably the best option for this migration could be configuring a cluster where against an Alfresco instance Solr is indexing and against the other instance Alfresco will be available for users (using Lucene).

When the indexing process is finished you can change the search subsystem to Solr and open the related instance, previously dedicated to Solr, to users.

Probably Alfresco can help you giving a limited Cluster license dedicated to perform this migration, please try to mention this idea to the Alfresco Support

Hope this helps.

[prim]

Thank you very much, PierGiorgio, I'll try your suggestion with Alfresco support.

Maybe we meet in this forum in the near future, after leaving the Enterprise world.

Bye

[mrogers]

The time taken to index is too long.   I agree with the suggestion to cluster alfresco, but there is something else going on which is why I suggested you contact support.  

For a start you don't say what sort of "objects" you are indexing.   That may be significant.   As may the number of docs to index.

[prim]

sort of "objects" we are indexing:

30% pdf readable, average 50 Kb
40% txt, doc, xls, average 900 Kb
10% images, average 2,000 KB
20% pdf non readable, average 3,210 Kb