zookeeper(11)源码分析-请求处理链(1)-创新互联

对于请求处理链而言,所有请求处理器的父接口为RequestProcessor。

创新互联是工信部颁发资质IDC服务器商,为用户提供优质的服务器托管雅安服务

RequestProcessor内部类RequestProcessorException,用来表示处理过程中的出现的异常,而proceequest和shutdown方法则是核心方法,是子类必须要实现的方法,处理的主要逻辑在proceequest中,通过proce***equest方法可以将请求传递到下个处理器。而shutdown表示关闭处理器,其意味着该处理器要关闭和其他处理器的连接。

public interface RequestProcessor {
    @SuppressWarnings("serial")
    public static class RequestProcessorException extends Exception {
        public RequestProcessorException(String msg, Throwable t) {
            super(msg, t);
        }
    }

    void proce***equest(Request request) throws RequestProcessorException;

    void shutdown();
}

实现RequestProcessor的processor有很多,PrepRequestProcessor,通常是请求处理链的第一个处理器。

PrepRequestProcessor

1、类的定义

public class PrepRequestProcessor extends ZooKeeperCriticalThread implements RequestProcessor {}

PrepRequestProcessor继承了ZooKeeperCriticalThread类并实现了RequestProcessor接口,表示其可以作为线程使用。

2、类核心成员

//已提交的请求队列
    LinkedBlockingQueue submittedRequests = new LinkedBlockingQueue();
    //下一个处理器
    private final RequestProcessor nextProcessor;
    // zk服务器
    ZooKeeperServer zks;

3、核心函数

3.1、run

while (true) {
                //从队列获取请求
                Request request = submittedRequests.take();
                long traceMask = ZooTrace.CLIENT_REQUEST_TRACE_MASK;
                if (request.type == OpCode.ping) {
                    traceMask = ZooTrace.CLIENT_PING_TRACE_MASK;
                }
                if (LOG.isTraceEnabled()) {
                    ZooTrace.logRequest(LOG, traceMask, 'P', request, "");
                }
                //requestOfDeath类型的请求,代表当前处理器已经关闭,不再处理请求。
                if (Request.requestOfDeath == request) {
                    break;
                }
                //调用关键函数
                pRequest(request);
            }

3.2、pRequest

pRequest会确定请求类型,并根据请求类型不同生成不同的请求对象,我们以创建节点为例子分析

//设置消息头和事务为空
        request.setHdr(null);
        request.setTxn(null);

        try {
            switch (request.type) {
            case OpCode.createContainer:
            case OpCode.create:
            case OpCode.create2:
                //创建节点请求
                CreateRequest create2Request = new CreateRequest();
                //处理请求
                pRequest2Txn(request.type, zks.getNextZxid(), request, create2Request, true);
                break;

                                //省略其他代码
                 //给请求的zxid赋值               
                 request.zxid = zks.getZxid();
                 //交给下一个处理器继续处理
         nextProcessor.proce***equest(request);

pRequest2Txn函数是实际的处理请求的函数,对于创建方法会调用pRequest2TxnCreate函数

//设置请求头
request.setHdr(new TxnHeader(request.sessionId, request.cxid, zxid,
                Time.currentWallTime(), type));

        switch (type) {
            case OpCode.create:
            case OpCode.create2:
            case OpCode.createTTL:
            case OpCode.createContainer: {
                pRequest2TxnCreate(type, request, record, deserialize);
                break;
            }

pRequest2TxnCreate方法如下:

private void pRequest2TxnCreate(int type, Request request, Record record, boolean deserialize) throws IOException, KeeperException {
        if (deserialize) {
            //反序列化,将ByteBuffer转化为Record
            ByteBufferInputStream.byteBuffer2Record(request.request, record);
        }

        int flags;
        String path;
        List acl;
        byte[] data;
        long ttl;
        if (type == OpCode.createTTL) {
            CreateTTLRequest createTtlRequest = (CreateTTLRequest)record;
            flags = createTtlRequest.getFlags();
            path = createTtlRequest.getPath();
            acl = createTtlRequest.getAcl();
            data = createTtlRequest.getData();
            ttl = createTtlRequest.getTtl();
        } else {
            //转换createRequest对象
            CreateRequest createRequest = (CreateRequest)record;
            flags = createRequest.getFlags();
            path = createRequest.getPath();
            acl = createRequest.getAcl();
            data = createRequest.getData();
            ttl = -1;
        }
        CreateMode createMode = CreateMode.fromFlag(flags);
        validateCreateRequest(path, createMode, request, ttl);
        //获取父节点路径
        String parentPath = validatePathForCreate(path, request.sessionId);

        List listACL = fixupACL(path, request.authInfo, acl);
        //获取父节点的record
        ChangeRecord parentRecord = getRecordForPath(parentPath);
        //检查ACL列表
        checkACL(zks, parentRecord.acl, ZooDefs.Perms.CREATE, request.authInfo);
        int parentCVersion = parentRecord.stat.getCversion();
        //是否创建顺序节点
        if (createMode.isSequential()) {
            //子路径后追加一串数字,顺序的
            path = path + String.format(Locale.ENGLISH, "%010d", parentCVersion);
        }
        validatePath(path, request.sessionId);
        try {
            if (getRecordForPath(path) != null) {
                throw new KeeperException.NodeExistsException(path);
            }
        } catch (KeeperException.NoNodeException e) {
            // ignore this one
        }
        boolean ephemeralParent = EphemeralType.get(parentRecord.stat.getEphemeralOwner()) == EphemeralType.NORMAL;
        //父节点不能是临时节点
        if (ephemeralParent) {
            throw new KeeperException.NoChildrenForEphemeralsException(path);
        }
        //新的子节点版本号
        int newCversion = parentRecord.stat.getCversion()+1;
        //新生事务
        if (type == OpCode.createContainer) {
            request.setTxn(new CreateContainerTxn(path, data, listACL, newCversion));
        } else if (type == OpCode.createTTL) {
            request.setTxn(new CreateTTLTxn(path, data, listACL, newCversion, ttl));
        } else {
            request.setTxn(new CreateTxn(path, data, listACL, createMode.isEphemeral(),
                    newCversion));
        }
        StatPersisted s = new StatPersisted();
        if (createMode.isEphemeral()) { //是否临时节点
            s.setEphemeralOwner(request.sessionId);
        }
        //拷贝
        parentRecord = parentRecord.duplicate(request.getHdr().getZxid());
        //子节点数量+1
        parentRecord.childCount++;
        //设置新版本号
        parentRecord.stat.setCversion(newCversion);
        //将parentRecord添加至outstandingChanges和outstandingChangesForPath中
        addChangeRecord(parentRecord);
        // 将新生成的ChangeRecord(包含了StatPersisted信息)添加至outstandingChanges和outstandingChangesForPath中
        addChangeRecord(new ChangeRecord(request.getHdr().getZxid(), path, s, 0, listACL));
    }

addChangeRecord函数将ChangeRecord添加至ZooKeeperServer的outstandingChanges和outstandingChangesForPath中。

private void addChangeRecord(ChangeRecord c) {
        synchronized (zks.outstandingChanges) {
            zks.outstandingChanges.add(c);
            zks.outstandingChangesForPath.put(c.path, c);
        }
    }

outstandingChanges 位于ZooKeeperServer 中,用于存放刚进行更改还没有同步到ZKDatabase中的节点信息。

znode节点会由于用户的读写操作频繁发生变化,为了提升数据的访问效率,ZooKeeper中有一个三层的数据缓冲层用于存放节点数据。

outstandingChanges->ZKDatabase->FileSnap+FileTxnLog

创新互联www.cdcxhl.cn,专业提供香港、美国云服务器,动态BGP最优骨干路由自动选择,持续稳定高效的网络助力业务部署。公司持有工信部办法的idc、isp许可证, 机房独有T级流量清洗系统配攻击溯源,准确进行流量调度,确保服务器高可用性。佳节活动现已开启,新人活动云服务器买多久送多久。


分享名称:zookeeper(11)源码分析-请求处理链(1)-创新互联
网页路径:http://ybzwz.com/article/dpepes.html