[ https://issues.apache.org/jira/browse/SOLR-565?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12613370#action_12613370 ]
patrick o'leary commented on SOLR-565: -------------------------------------- Been tied up with work recently The objective of this is to move shards away from a client implementation, with the nature of solr I would expect engineer resources to be used for solrconfig.xml updates, but operational updates to occur externally to this, hence hosting topology externally to the solrconfig. The distributor component should end up providing 3 main parts, that don't really fit into the xml config. 1. Component with the ability to enable shard-ing based upon a distributed=true parameter from a preset topology. 1.1 Overridden if isShard=true is passed. (thus local disk only) 2. a configurable plugin to allow loading of custom topologies, e.g. mysql, ec2-describe-instances, zoneedit-watch etc. 2.1. defaulting to use a standard file loader, which checks file mod time stamp for reload. 3. provide a configurable plugin for query routing, if users wish to apply taxonomies or other query identification methods, they can do so and direct requests to appropriate shards. I don't think this really is needed as part of solr's core, and if anything could be a contrib. > Component to abstract shards from clients > ----------------------------------------- > > Key: SOLR-565 > URL: https://issues.apache.org/jira/browse/SOLR-565 > Project: Solr > Issue Type: New Feature > Components: search > Affects Versions: 1.3 > Reporter: patrick o'leary > Assignee: Otis Gospodnetic > Priority: Minor > Attachments: distributor_component.patch > > > A component that will remove the need for calling clients to provide the > shards parameter for > a distributed search. > As systems grow, it's better to manage shards with in solr, rather than > managing each client. -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online.