Solr4.8.0源码分析(2)之Solr的启动(一)

cnn

　　上文写到Solr的启动过程是在SolrDispatchFilter的init()里实现，当Tomcat启动时候会自动调用init();
　　Solr的启动主要在 this.cores = createCoreContainer();语句中实现。



/**
*初始化，当tomcat启动时候开始初始化，其中主要调用createCoreContainer来实现Solr的初始化
*/
public void init(FilterConfig config) throws ServletException
{
log.info("SolrDispatchFilter.init()");
try {
// web.xml configuration
this.pathPrefix = config.getInitParameter( "path-prefix" );
this.cores = createCoreContainer();
log.info("user.dir=" + System.getProperty("user.dir"));
}
catch( Throwable t ) {
// catch this so our filter still works
log.error( "Could not start Solr. Check solr/home property and the logs");
SolrCore.log( t );
if (t instanceof Error) {
throw (Error) t;
}
}
log.info("SolrDispatchFilter.init() done");
}

　　Solr的启动过程主要包括：
　　1. 获取SolrHome：分别先后通过JNDI，System property，default directory三种方式尝试获取
2. 实例化启动过程中使用的类加载器SolrResourceLoader
3. 加载solrhome下的solr.xml文件，封装为ConfigSolr
4. 实例化一个CoreContainer,通过CoreContainer来加载cores
5. 遍历SolrHome，当寻找到含有core.properties的文件夹，则作为一个core
6. 多线程加载cores
7. 加载每个core时先加载solrconfig.xml封装为SolrConfig
8. 再加载schema.xml封装为IndexSchema
9. 最后实例化SolrCore
　　以下是createCoreContainer()实现代码，该函数就包含了以上的启动过程。



1   /**
2    * Override this to change CoreContainer initialization
3    * @return a CoreContainer to hold this server's cores
4    */
5   protected CoreContainer createCoreContainer() {
6     SolrResourceLoader loader = new SolrResourceLoader(SolrResourceLoader.locateSolrHome());
7     ConfigSolr config = loadConfigSolr(loader);
8     CoreContainer cores = new CoreContainer(loader, config);
9     cores.load();
10     return cores;
11   }
　　1. 获取SolrHome信息
　　Solr通过三种方式来获取SolrHome,分别是JNDI,System property,以及当前路径。
　　JNDI的java:comp/env/solr/home的获取，主要通过在tomcat/conf/Catalina/localhost目录下新建solr.xml。Tomcat启动的时候就会该目录下自动获取  SolrHome的信息.
　　System property的SolrHome获取是直接在tomcat启动脚本catalina.sh中加入set JAVA_OPTS -Dsolr.solr.home=/Users/rcf/workspace/java/solr 。
　　最后的获取SolrHome的方式就是在当前目录下查找。



1   /**
2    * Determines the solrhome from the environment.
3    * Tries JNDI (java:comp/env/solr/home) then system property (solr.solr.home);
4    * if both fail, defaults to solr/
5    * @return the instance directory name
6    */
7   /**
8    * Finds the solrhome based on looking up the value in one of three places:
9    *
10    *  JNDI: via java:comp/env/solr/home
11    *  The system property solr.solr.home
12    *  Look in the current working directory for a solr/ directory
13    *
14    *
15    * The return value is normalized.  Normalization essentially means it ends in a trailing slash.
16    * @return A normalized solrhome
17    * @see #normalizeDir(String)
18    */
19   public static String locateSolrHome() {
20
21     String home = null;
22     // Try JNDI
23     try {
24       Context c = new InitialContext();
25       home = (String)c.lookup("java:comp/env/"+project+"/home");
26       log.info("Using JNDI solr.home: "+home );
27     } catch (NoInitialContextException e) {
28       log.info("JNDI not configured for "+project+" (NoInitialContextEx)");
29     } catch (NamingException e) {
30       log.info("No /"+project+"/home in JNDI");
31     } catch( RuntimeException ex ) {
32       log.warn("Odd RuntimeException while testing for JNDI: " + ex.getMessage());
33     }
34     
35     // Now try system property
36     if( home == null ) {
37       String prop = project + ".solr.home";
38       home = System.getProperty(prop);
39       if( home != null ) {
40         log.info("using system property "+prop+": " + home );
41       }
42     }
43     
44     // if all else fails, try
45     if( home == null ) {
46       home = project + '/';
47       log.info(project + " home defaulted to '" + home + "' (could not find system property or JNDI)");
48     }
49     return normalizeDir( home );
50   }


1  
2     
3   
　　获取完SolrHome后,就会实例化SolrResourceLoader。因为传进去的parent是null，所以parents是ContextClassLoader，并且加载了SolrHome/lib/下的文件.



1   /**
2    *
3    * This loader will delegate to the context classloader when possible,
4    * otherwise it will attempt to resolve resources using any jar files
5    * found in the "lib/" directory in the specified instance directory.
6    *
7    *
8    * @param instanceDir - base directory for this resource loader, if null locateSolrHome() will be used.
9    * @see #locateSolrHome
10    */
11   public SolrResourceLoader( String instanceDir, ClassLoader parent, Properties coreProperties )
12   {
13     if( instanceDir == null ) {
14       this.instanceDir = SolrResourceLoader.locateSolrHome();
15       log.info("new SolrResourceLoader for deduced Solr Home: '{}'",
16                this.instanceDir);
17     } else{
18       this.instanceDir = normalizeDir(instanceDir);
19       log.info("new SolrResourceLoader for directory: '{}'",
20                this.instanceDir);
21     }
22     
23     this.classLoader = createClassLoader(null, parent);
24     addToClassLoader("./lib/", null, true);
25     reloadLuceneSPI();
26     this.coreProperties = coreProperties;
27   }


1   /**
2    * Convenience method for getting a new ClassLoader using all files found
3    * in the specified lib directory.
4    */
5   static URLClassLoader createClassLoader(final File libDir, ClassLoader parent) {
6     if ( null == parent ) {
7       parent = Thread.currentThread().getContextClassLoader();
8     }
9     return replaceClassLoader(URLClassLoader.newInstance(new URL[0], parent),
10                               libDir, null);
11   }



1   /**
2    * Adds every file/dir found in the baseDir which passes the specified Filter
3    * to the ClassLoader used by this ResourceLoader.  This method MUST
4    * only be called prior to using this ResourceLoader to get any resources, otherwise
5    * it's behavior will be non-deterministic. You also have to {link @reloadLuceneSPI}
6    * before using this ResourceLoader.
7    *
8    * This method will quietly ignore missing or non-directory baseDir
9    *  folder.
10    *
11    * @param baseDir base directory whose children (either jars or directories of
12    *                classes) will be in the classpath, will be resolved relative
13    *                the instance dir.
14    * @param filter The filter files must satisfy, if null all files will be accepted.
15    * @param quiet  Be quiet if baseDir does not point to a directory or if no file is
16    *               left after applying the filter.
17    */
18   void addToClassLoader(final String baseDir, final FileFilter filter, boolean quiet) {
19     File base = FileUtils.resolvePath(new File(getInstanceDir()), baseDir);
20     if (base != null && base.exists() && base.isDirectory()) {
21       File[] files = base.listFiles(filter);
22       if (files == null || files.length == 0) {
23         if (!quiet) {
24           log.warn("No files added to classloader from lib: "
25                    + baseDir + " (resolved as: " + base.getAbsolutePath() + ").");
26         }
27       } else {
28         this.classLoader = replaceClassLoader(classLoader, base, filter);
29       }
30     } else {
31       if (!quiet) {
32         log.warn("Can't find (or read) directory to add to classloader: "
33             + baseDir + " (resolved as: " + base.getAbsolutePath() + ").");
34       }
35     }
36   }