循序渐进企业级搜索引擎之Master模块设计及实现-系统请求响应封装模块

2.2 系统请求响应封装模块

 

系统首先从request的信息中获取请求的类型，并依据请求的类型获取相应的处理类，

if (handler == null && path.length() > 1) {
handler = HandlerFactory.getHandler(path);
}

 

//依据请求的命令来获取响应的处理类
public static ChameleonRequestHandler getHandler(String cmd) {
ChameleonRequestHandler handler = null;
if (cmd.equalsIgnoreCase(ChameleonConstants.INDEX) ||
cmd.equalsIgnoreCase(ChameleonConstants.UPDATE) ||
cmd.equalsIgnoreCase(ChameleonConstants.CRAWLER) ||
cmd.equalsIgnoreCase(ChameleonConstants.DELETEQUERY) ||
cmd.equalsIgnoreCase(ChameleonConstants.DELETETERM))
handler = (ChameleonRequestHandler) WebAppContextUtil
.getBean("indexRequestHandler");
…
return handler;
}

接下来就是对request的处理，首先是一个基类处理如下：

public abstract class RequestHandlerBase implements ChameleonRequestHandler{

@Override
public void handle(ChameleonQueryRequest req, ChameleonQueryResponse rsp) {
startHandle(req,rsp);
}

private void startHandle(ChameleonQueryRequest req, ChameleonQueryResponse rsp){
handleResponse(rsp,handleRequest(req));
}

public abstract String handleRequest(ChameleonQueryRequest req);

public abstract void handleResponse(ChameleonQueryResponse rsp,String result);

}
@Override
public void handle(ChameleonQueryRequest req, ChameleonQueryResponse rsp) {
startHandle(req,rsp);
}

private void startHandle(ChameleonQueryRequest req, ChameleonQueryResponse rsp){
handleResponse(rsp,handleRequest(req));
}

public abstract String handleRequest(ChameleonQueryRequest req);

public abstract void handleResponse(ChameleonQueryResponse rsp,String result);

 
每个handler都是对RequestHandlerBase扩展，需要扩展的是handleRequest和handleResponse
具体处理index的类如下：

 

@Override
public String handleRequest(ChameleonQueryRequest req) {
MasterCfgManager.getIndexPool().execute(new IndexTask(req));
return null;
}

@Override
public void handleResponse(ChameleonQueryResponse rsp, String code) {
//to do
}

 

在系统index线程池中放入一个indextask，该线程池初始化的大小是系统cpu的个数，具体的excute执行如下代码：

@Override
public void excuteProvider(ChameleonQueryRequest cqr) throws BaseException {
ZooKeeperClientInfo client = CalculateLoadUtil
.selector(CalculateServerType.ALL);
ChameleonExcute ce = SlaveClientFactory.getInstance("", "")
.createClient(ChameleonExcute.class, client.getIp());
ce.parallelMultiIndex(cqr.getBody());
}

 

首先依据slaves的负载情况选择一个处理的客户端，并依据此客户端获取执行实例ChameleonExcute，接下来就是请求到客户端处理实际的索引。而CalculateLoadUtil就是系统处理负载均衡模块的入口，接下来具体讲解。