[jira] [Comment Edited] (SOLR-7073) Add an API to add a jar to a collection's classpath

["Noble Paul (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SOLR-7073?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14343469#comment-14343469 ] 

Noble Paul edited comment on SOLR-7073 at 3/2/15 6:05 PM:
----------------------------------------------------------

This does a complete refactor of Solr's plugin loading because all components now need to be lazily loadable based on the extra flag. I would love to have extra eyes on this before it is too late. 

There is a new class called {{PluginRegistry}} which is slightly more than our {{Map<String,Plugin>>}} .{{PluginRegistry}} takes care of loading the plugins at startup or lazily depending on the various attributes such as {{startup="lazy"}} or {{runtimeLib="true"}} . This patch gets rid of the Lazy loading wrappers we had for {{SolrRequestHandler}} and {{QueryResponseWriter}}


was (Author: noble.paul):
This does a complete refactor of Solr's plugin loading because all components now need to be lazily loadable based on the extra flag. I would love to have extra eyes on this before it is too late. 

> Add an API to add a jar to a collection's classpath
> ---------------------------------------------------
>
>                 Key: SOLR-7073
>                 URL: https://issues.apache.org/jira/browse/SOLR-7073
>             Project: Solr
>          Issue Type: Sub-task
>            Reporter: Noble Paul
>            Assignee: Noble Paul
>         Attachments: SOLR-7073.patch
>
>
> The idea of having separate classloaders for each component may be counter productive.  This aims to add a jar dependency to the collection itself and each core belonging to that collection will have the jar in the classpath
> As we load everything from the .system collection , we cannot make the core loading delayed till .system is fully loaded and is available . 
> There is a new  set of  config commands to manage the dependencies on a collection level. It is possible to have more than one jar as a dependency. The "lib" attribute is same as the blob name that we use in the blob store API
> example:
> {code}
> curl http://localhost:8983/solr/collection1/config -H 'Content-type:application/json'  -d '{
> "add-runtimelib" : {"name": "jarname" , "version":2 },
> "update-runtimelib" :{"name": "jarname" ,"version":3},
> "delete-runtimelib" :"jarname" 
> }' 
> {code}
> The same is added to the overlay.json .
> The default SolrResourceLoader does not have visibility to  these jars . There is a classloader that can access these jars which is made available only to those components which are specially annotated
> Every pluggable component can have an optional extra attribute called {{runtimeLib=true}}, which means, these components are not be loaded at core load time. They are tried to be loaded on demand and if all the dependency jars are not available at the component load time an error is thrown . 
> example of registering a valueSourceParser which depends on the runtime classloader
> {code}
> curl http://localhost:8983/solr/collection1/config -H 'Content-type:application/json'  -d '{
> "create-valuesourceparser" : {"name": "nvl" ,
> "runtimeLib": true, 
> "class":"solr.org.apache.solr.search.function.NvlValueSourceParser , 
> "nvlFloatValue":0.0 }  
> }'
> {code} 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org