On 5/13/2015 1:03 AM, Guy Thomas wrote: > Up to now we’ve been using Lucene without Solr. > > The Lucene index is being updated and when the update is finished we > notify a Hessian proxy service running on the web server that wants to > read the index. When this proxy service is notified, the server knows it > can read the updated index. > > Do we have the use a similar set-up when using Solr, that is: > > 1. Create/update the index > > 2. Notify the Solr client
In Solr, the Solr server has complete control of the Lucene index and maintains the write lock at all times. Generally you create or update the index via requests to Solr, through the update handler. As soon as you issue a commit with openSearcher=true, and it completes, all clients can see the changes. There is no need to do any kind of notification. Commits may be fully automated within the Solr configuration or they may be explicitly sent by clients. If you are creating the index in some other way, then you generally need to reload the core. Recently (5.x versions) at least one person has been having trouble with loading a new index using RELOAD: https://issues.apache.org/jira/browse/SOLR-7526 Thanks, Shawn
